Entire copy skill stack

Read-only audit snapshot. These are the full source files, including older rules and amendments. Nothing has been reconciled, shortened, or rewritten. Inclusion is not endorsement of every rule or claim. This is a public review page.

44 complete Markdown files. Native-copy plus its references/examples, strategy, humanizer, QA, and core brand doctrine. Other external documents linked inside these files are not recursively included.

Download original source bundle · Use your browser’s Find to search all expanded text.

Native copy / SKILL.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/SKILL.md

SHA-256: 65a1a2e375cd6cae2df2457bafd60cda2f7fa46c8978277a1aeb54234b0c84a6

Full source · 657 lines
0001  ---
0002  name: native-copy
0003  description: Write native ad copy variants for DawnBands (or other brands). Use when creating Facebook/Instagram native ad copy, writing ad copy variants for A/B testing, generating headlines + body copy for Meta ads, or writing copy for native ads pointing to listicle advertorials. Triggers on "write native ads", "ad copy variants", "write copies for [ad]", "native copy", "Meta ad copy", "5 variants for [ad]".
0004  ---
0005  
0006  # Native Copy Skill
0007  
0008  Write high-converting native Facebook ad copy using Zakariah's framework, deep avatar research, and proven format rules.
0009  
0010  
0011  
0012  ## IMAGE LANE ORDER (Fish 2026-08-25 — PERMANENT)
0013  
0014  1. RIP FIRST: pull_native_rips.py with queries shaped like real post titles ("my son's room",
0015     "light under the door"), not staged ad-scene descriptions. Review every candidate by eye.
0016  2. A real rip that passes gates TAKES THE SLOT over any gen, even when a gen matches the
0017     copy's scene better. Nativeness > relevance is absolute (Fish receipt: charging-caddy rip
0018     over on-scene kitchen gen, 8-25).
0019  3. Gen fallback is NEVER bare-prompt. Pull a real Pinterest/web reference photo FIRST
0020     (scrapling + curl, the boring amateur photo not the good one, demographic-matched when
0021     people appear) and pass it as the style reference. Fake-UI/help-post recreations reference
0022     a real screenshot of the actual UI.
0023  
0024  ## DIRECTNESS — THE THREE-STRIKE ERROR (Fish 2026-09-01, PERMANENT)
0025  
0026  Fish rejected three consecutive rewrites of the same passage. Each fixed the surface and kept
0027  the literary move. Check all three before shipping any line.
0028  
0029  1. **ROUNDABOUT** — "What I ordered was the first knock... the honest answer to what happens
0030     when he leaves home is that I am the answer."
0031  2. **ANAPHORA** (only *sounds* direct) — "To stop making the first knock. / To stop walking to
0032     his door. / To stop hearing my own voice before seven." Three or more consecutive lines
0033     opening with the same word or phrase is a poem. Fish: "make sure people actually say that
0034     ... or itll throw them off."
0035  3. **ROLE-ABSTRACTION** — "right now I am still the plan." Fish: *"i am still the plan is
0036     still not what a mom says."* She never becomes an abstract noun.
0037  
0038  **GOOD:** "I am tired of being his alarm clock. I do not want to start every day angry at my
0039  own kid. He leaves next year, and I will still be the one calling him every morning."
0040  
0041  Banned: `I am still the plan` · `I am the answer` · `that was the variable` · `never in that
0042  channel` · `that is the whole thing` · `the part I could not fix was` · `What I ordered was X`
0043  
0044  Instead: name the concrete action she performs, or the feeling she has. Register anchor from
0045  real buyers: "Want to start the school year off with out the fighting." / "Hoping it stops
0046  angry mornings I wake up in fight or flight mode."
0047  
0048  Exception: real idioms mothers use are fine — "I am his alarm clock", "human snooze button".
0049  
0050  Also binding: grade 3-5, average sentence under 14 words, nothing over 28, active voice only,
0051  say HABITUATES / TUNES IT OUT / AUTOMATICALLY in plain words.
0052  
0053  **Readability is semantic, not only numerical (Fish 2026-09-04, permanent).** A short sentence can still be nonsense. If a line needs translation, describes a mental state as an abstract status, or sounds like compressed copywriter logic, it fails even when the grade and sentence-length checks pass. Ban constructions such as `just not started`, `getting up earlier held until nothing moved`, and `your calm is his calm`. Name what her son has or has not physically done and what she literally thought: `He stood beside his shoes without reaching for them.` The final read includes the brief’s readability target, average sentence under 14, zero sentences over 28, and this said-out-loud sense check.
0054  
0055  **Readability targets are brief-scoped.** Default remains grade 3–5 unless the active batch tightens it. When Fish names a tighter band (for example grade 2–4), put that range in the structured brief and deterministic gate for that batch; do not silently treat a session-specific target as permission to loosen or tighten unrelated formats.
0056  
0057  **Failed-solution paragraphs must be linear and direct (Fish 2026-09-04, permanent).** One attempt per beat, in this order: what she tried → why she tried it → what literally happened → why she stopped trusting it. Never jumble several methods into compressed fragments or slogan summaries. Ban `I tried everything the posts say`, `came next`, `laid it on`, `used it up`, `both helped`, and `neither stopped`. A reader should never have to reconstruct the timeline. Example: `I started getting up at 5:30 because I thought I was the problem. I wanted to be ready and calm before he got up. I still ended up yelling by the door.` Full operator card: `references/direct-failed-solution-prose.md`.
0058  
0059  **No coffee anywhere (Fish 2026-09-04, permanent).** Do not use coffee as morning texture, a reaction beat, a warm/cold payoff, or filler in any native. Use a story-owned action or object instead.
0060  
0061  **Meta-lesson:** a fix that changes syntax but keeps the literary SHAPE is not a fix.
0062  Directness is not compression. It is plain speech.
0063  
0064  Full doctrine: `references/copy-directness-doctrine.md` · enforced by
0065  `~/.openclaw/workspace/scripts/validate_native_copy.py`
0066  
0067  ## RELATABILITY GATE — no unverified qualifiers (Fish 2026-09-03, PERMANENT)
0068  
0069  Fish, on the SA2 guilt native audit: "Relatability risk audit needs to be part of the written
0070  pipeline where no need to add extra qualifiers that we can't verify. Subavatars are only tied by
0071  the specific categories that we chose."
0072  
0073  - The reader is identified ONLY through the sub-avatar card's categories (one experience, one
0074    emotion, the behaviors that follow). Every other detail is a qualifier, and a qualifier the
0075    reader does not share is a mother lost.
0076  - Every identification line carries a receipt: a quoted commenter or a survey verbatim in the
0077    brief. No receipt, no line. Invented items (water bottle, lunch box), invented kid temperament
0078    ("he always forgives me"), invented places (the store, dinner), clock times ("6:50", "before
0079    seven"), ages, grades, siblings, partners, a car or bus as the reader's morning: all out unless
0080    the card carries them.
0081  - Her own process is written the way the mothers state it, not compressed for rhythm: she says
0082    it over and over, nothing moves, the voice goes up, the guilt after. No count on the repeats,
0083    ever, even when the card's quote carries one ("at least twenty" is one mother's number, not the
0084    reader's). "Loud by the third reminder" and "twenty reminders" were the 9-03 misses.
0085  - The mechanism is a stated CONTRAST, never a feature line: sound fills the room and startles,
0086    a parent at the door is an order on a half-asleep kid, both spike stress hormones; the wrist
0087    is small, quiet, touches only him, gives no order, he ends it himself, so no cortisol spike,
0088    clearer head, calmer nervous system. If a reader cannot say WHY the wrist is different after
0089    one read, the mechanism is under-stated (Fish 9-03).
0090  - Owned objects must be real situations from the source research (the toothbrush with the paste on it,
0091    dry), never a staged prop (the sticky note that says "don't yell" was rejected 9-03).
0092  - **Specific failed solutions are not universal qualifiers (Fish 2026-09-04, permanent).** Visual timers,
0093    charts, rainbow clocks, getting up earlier, staying in the room, and other methods may appear only as
0094    this narrator's receipted history. Never put any of them in the reader bridge or imply every SA2 mother
0095    tried them. The reader is identified by the 1:1:1 combination only: repeated instructions until her
0096    voice rises, guilt, and apologizing after. Vary the narrator's failed-solution stack rather than cloning
0097    one master checklist across bodies.
0098  - **Failed-solution restraint + whole-morning mechanism (Fish 2026-09-04, permanent).** Never turn a native
0099    into an alarm-category catalog. For the ADHD teen-mom lane, use at most three prevalence-weighted
0100    families, in this order when all are present: (1) Mom waking him herself, (2) phone at full volume plus
0101    loud alarm clocks grouped into ONE sound family, (3) routines / bedtimes / sleep apps. Each family gets
0102    acknowledge → gap → one memorable summary in a sympathetic, educational tone. Smartwatch/fitness band,
0103    bed/pillow shaker, medication, “something else,” and “nothing yet” never become standalone lectures;
0104    use one only when the narrator or objection specifically requires it, and keep tactile failures out when
0105    they confound our mechanism. The mechanism is not “our alarm is better.” Explain the whole chain: sound
0106    fails → Mom becomes the backup → his first interaction is her order while half awake → repetition and
0107    pushback shape the next hour → a private teen-owned wrist cue removes Mom from the first step. Preserve
0108    the honest limit: the band changes how the morning begins and who owns its first action; it does not claim
0109    to erase every later executive-function problem.
0110  - **Body doubling is real but narrow.** We have a direct parent VOC receipt: "Before I started body
0111    doubling, it was just all yelling." It is not universal. Prefer the concrete action (`I stayed in the
0112    room with him for every step`) unless the exact speaker uses the term.
0113  - **"Connection over correction" is not an exact phrase in the Dawn corpus.** The underlying action of
0114    sitting with the child and trying to connect is receipted, but the slogan is not. Describe the action;
0115    do not present the label as a standard ADHD method.
0116  - **Consumer-facing source channel for Meta natives: Facebook `post/posts` only (Fish 2026-09-04).**
0117    Do not alternate among thread, forum, and post inside an ad. Internal provenance may still identify a
0118    Reddit thread or forum URL; that source taxonomy never leaks into the narrator's words.
0119  - Narrator-only devices (an owned object, a named person replying to her post) stay hers; never phrase them
0120    as something the reader has too. Authority first names may vary by body; the character's controlled
0121    credentials and mechanism function remain fixed unless the brief changes them.
0122  - BRIEF GATE (Fish 2026-09-04, PERMANENT: "how can we make sure the briefs we write follow our native
0123    rules, do not add overqualifications, follow 1:1:1, do not hallucinate things that happen and use
0124    VOC"). The body validator catches what the model added; the brief gate catches what I added, and it
0125    runs BEFORE any generation: `python3 ~/.openclaw/workspace/scripts/brief_gate.py <units.json>`
0126    (exit 1 = no run). The unit brief is structured, not prose: every scene field carries a receipt of
0127    class `voc` (verbatim substring of the corpus, thread id), `narrator` (her identity texture only)
0128    or `dramatized` (allowed ONLY in the sanctioned slots: kid lines with a corpus-hit `shaped_from`,
0129    one noticer line, the discovery chain, Dana's analogy, counted results). FAILs: card not 1:1:1 or
0130    desire not a mass desire; any field without a receipt; a paraphrase presented as a quote; a
0131    dramatized field outside its slot; qualifier scan on reader-facing fields (clock time, age/grade,
0132    count on repeats, reader's car/bus, meds/screen/bedtime/app, the 0-hit my-phrase list); sibling /
0133    partner / bus anywhere; figurative constructions in any field (simile, "handing him her own",
0134    labor metaphor, swap/mirror, personified morning, coined event label, the named strikes: builds,
0135    braced, flash of, lose the thread, leftover, the other one); qualifier nouns (water bottle, lunch
0136    box, hair, uniform, checklist, sticky note); any content word in a scene field with 0 corpus hits
0137    on every stem (my-word-in-her-slot, 7 strikes); a person noun with 0 hits; the system brief
0138    missing any exact required line (Cortese, company intro naming the brand, guarantee line,
0139    clearer head, nervous system, gives up, not a phone, the close rungs). The shared body BANNED list
0140    is imported so a same-day ruling covers briefs the same turn. Model batch:
0141    output/dawn-sa2-guilt-tof-2026-09-03 (brief-units-mid-angles.json, audit-brief-*.md).
0142  - Enforced: `scripts/validate_native_copy.py --voc <brief>` flags clock times, qualifier nouns,
0143    and identification lines with no content-word overlap with the brief's quoted material. Run
0144    it before qa-gate; a flagged line is either receipted in the brief or cut. The line-by-line
0145    receipt table is written to `audit-1-1-1.md` in the batch dir (model: dawn-sa2-guilt-tof-2026-09-03).
0146  
0147  ## GATES, IN ORDER — every long-format body, every time (Fish 2026-09-04: "make sure these passes are always being ran for future copies")
0148  
0149  One command runs them and stops at the first failure:
0150  `python3 ~/.openclaw/workspace/scripts/native_gates.py --units <units.json> --bodies <final/*.txt> --band mid|long --objects <obj,obj,...> [--card <subavatar-card.json> --corpus <receipt-corpus.txt> --attempt-profile <failed-solution-profile.json>]`
0151  
0152  For non-default sub-avatars, the optional card/corpus/profile flags are mandatory: the runner forwards the
0153  selected card and receipt corpus into `brief_gate.py`, then forwards the attempt profile into `audit_sop.py`.
0154  The profile supports at most three ordered failed-solution families; aliases such as phone and loud alarms
0155  collapse into one family. Each family must contain acknowledge → gap → explicit memorable-summary evidence.
0156  Standalone low-frequency lanes, repeated fourth family spans, malformed profiles, grade outside 2–4,
0157  average sentences over 13 words, or a sentence over 28 words hard-fail. Stage 5 parses the receipt Markdown
0158  fail-closed, including escaped pipes, and requires `field | class | provenance | authorization`; dramatized
0159  week/count/result rows need an explicit sanctioned-slot authorization.
0160  
0161  1. `scripts/brief_gate.py` on the unit briefs (BEFORE generation, and again here so the shipped brief is the gated brief).
0162  2. The batch `validate.py` per body (band, owned object, offer floor, relatability regexes, shared BANNED) and once across all bodies for cloned lines. Mixed-object batch validation will report expected object misses because `validate.py` accepts one global object; `native_gates.py` treats the per-unit runs as authoritative and fails the convergence-only call only when `CLONED` lines are present.
0163  3. `scripts/audit_sop.py` — the SOP 2.0 mechanical pass: §10 formatting numbers, §04 part word-shares, §08 reveal %, §06 dialogue:reaction pacing, §07 stack items addressed after the mechanism, §09 six weeks + counted results, §03 analogy + uniqueness link. Hard fails: reveal outside 65-82, fewer than six weeks, no analogy, no uniqueness link, a product hint in the first third, any emoji but the two 👉, bold/caps.
0164  4. `audit-sop-<batch>.md` must exist: the §11 eight killers and §13 final check answered per body BY HAND with line receipts (killers 2 shallow avatar, 4 tone switch, 7 stakes beyond the individual, 8 weak incident, and §13 voice / specificity / keep-reading are judgment, not regex). Model: output/dawn-sa2-guilt-tof-2026-09-03/audit-sop-v6.md.
0165  5. `audit-1-1-1*.md` must exist: the line-by-line receipt table (scripts model: audit_v6.py in the same batch).
0166  Then qa-gate, then Fish. A body that has not been through all five is not "done".
0167  
0168  For incident-regeneration batches and Cloudflare review-page delivery, load `references/peak-incident-finalization-and-review-page.md`. It covers partial promotion from raw attempts, final-file marker checks, stale-audit repair, detector false-positive handling, believability-versus-polarization judgment, review-ready versus launch-ready status, and live production verification.
0169  
0170  When a reader flags foreign phrasing, confusing mechanism reactions, overqualified bridges, cloned failed-solution stacks, source-channel drift, or filler texture, load `references/readability-relatability-repair.md`. It separates direct VOC from dramatization and model synthesis, then gives the concrete-language repair and verification sequence. For the final paragraph-by-paragraph semantic gate after mechanical SOP validation, also use `references/directness-and-jumble-final-pass.md`; it catches unclear speaker turns, chronology, pronouns, compressed causality, and callbacks that numeric readability misses.
0171  
0172  INCITING INCIDENT (Fish 2026-09-04, PERMANENT): the incident is written at the PEAK OF BELIEVABILITY, not at the average
0173  morning, as long as everything before it (the world, the stack, the toll) is verified and believable so the reader has
0174  already decided the narrator is real when the peak comes. The OBJECT keeps its VOC receipt; the PEAK is a dramatized slot
0175  (`incident_peak` in the unit brief, receipted with the VOC lines it is shaped from). Visceral and physical, nobody new
0176  enters, nothing invented outside her, her son and the object. SOP §11 killer 8 is checked against this bar, not the
0177  average morning.
0178  
0179  ## SAID-OUT-LOUD TEST (Fish 2026-08-25 — PERMANENT, applies to every native)
0180  
0181  Every narrator line must read as something the narrator would literally say out loud to
0182  another mom. Proof source: the raw Reddit parent threads this lane is mined from — real
0183  moms state what happened, flat and direct, zero symbolism. Kill on sight:
0184  - composed parallel rhythm ("A fight at 11. A job at 6:45.")
0185  - announcer transitions ("Here is the part nobody warned me about.")
0186  - mirrored abstractions ("the phone moved in, the job moved onto me")
0187  - report language ("came with tension" -> say "turned into a thing")
0188  - analogies and similes ANYWHERE in body or hooks ("like I am arresting him")
0189  - symbolism of any kind; describe directly what is going on instead
0190  - coined event labels ("the 11 o'clock handover", "phones sleep in the kitchen") — name the literal action every time
0191  - personification ("the phone slept in my room") — objects charge, sit, go places; only people sleep. Exception: verbatim Reddit VOC quoted as-is
0192  Run this as a dedicated pass after drafting, separate from and in addition to the humanizer pass.
0193  
0194  **HOOK OPENING — TWO LEGAL FAMILIES (Fish 2026-08-31, amends the 8-25 first-person-only rule):**
0195  
0196  1. **First-person lived action** — "I/My/We" + concrete action. The 8-25 rule, unchanged. ("I used to make three trips up the stairs before 7am.")
0197  2. **Second-person symptom flag** — "If your [observable symptom]..." Fish 8-31: **"allow the 'if your' hook family, that's what converts."** Receipt: `adhd convenience slop 7` opens "If your ADHD teen 'can't' wake up to alarms, you're not dealing with a discipline problem" and is the account's biggest converter ($14,475 spend, 493 purchases, $29 CPA). Independently confirmed as the dominant device across SPNutrition's catalog (their #1 hook by impressions is a symptom flag).
0198  
0199  Everything else in the said-out-loud test still binds on BOTH families: no analogies, no composed parallel rhythm, no announcer transitions, no coined event labels, no personification. A flag hook still names a CONCRETE observable ("sleeps through every alarm", "has a psychiatrist, a therapist and a 504 plan"), never a vague state ("struggles in the morning").
0200  
0201  **Still banned:** the third family — pure mystery/curiosity openers that flag nobody. That is the unaware-pulling shape the awareness gate parks and the failure mode behind our expensive-click ads.
0202  
0203  ## DawnBands Selling Doctrine — Mandatory
0204  
0205  Before writing, briefing, editing, or auditing any DawnBands native, read:
0206  
0207  `/Users/ayden/.openclaw/workspace/knowledge/brands/dawnbands/selling-root-cause-mechanism.md`
0208  
0209  It is the canonical consumer-facing mechanism doctrine for every mode and overrides older claim-safe-only framing that reduces DawnBands to “a different cue placement.” Future DawnBands copy should use the full causal chain where format permits:
0210  
0211  `circadian delay → sleep debt → auditory gating → sleep inertia / half-awake automatic behavior → Mom becomes the emergency pathway → Somatic Wake Bypass™ → Wrist-First Handoff™ → relationship + independence payoff`
0212  
0213  Aggressive biological simplification is allowed and desired for the selling layer. Keep exact-SKU performance, testimonial timelines, offer facts, and customer outcomes tied to receipts; do not let provenance review flatten the root cause or mechanism into generic feature language.
0214  
0215  ## ⛔ AWARENESS SCOPE GATE — UNAWARE LANE PARKED (Fish 2026-08-31, binding until he lifts it)
0216  
0217  **DawnBands runs PROBLEM-AWARE and warmer only. Do not write, brief, or recommend unaware-audience copy.**
0218  
0219  Fish's reason: the market isn't there yet and we cannot scale at the CPA unaware traffic requires. This is a business-math ruling, not a craft judgment — the formats below are not "bad," they are unaffordable right now.
0220  
0221  **PARKED (do not use, do not propose):**
0222  - **T3 unaware study-reveal** — parked entirely (it was always the unaware template, and it had no spend receipt anyway; every stat/study-hook ad we launched starved).
0223  - **The unaware application of T1 / Mode B**, including the zero-category-scent cold opener strategy built for readers who don't know they have the problem.
0224  - **ONLY the unaware-specific device in the Alevia corpus: the zero-category-scent mundane-object cold open** ("mundane object + weekday + month + place-name in sentence one, zero product scent"). That opener is engineered for readers who do not know they have the problem, and it is exactly the artifact-open trap — it withholds the TARGET. Parked for our traffic.
0225  
0226  ⚠️ **DO NOT over-park Alevia (Fish 2026-08-31).** The audience is parked, the CRAFT is not. `alevia-live-anatomy-2026-08-12.md` is measured across 240 live copies and stays fully in force for problem-aware work. Keep using, unchanged:
0227  - **The P.S. ladder** — 217/240 of their copies close on stacked postscripts, and this file calls it "our biggest structural gap." Rungs each have a distinct job: P.S. = rupture-object return · P.P.S. = guarantee naturalized in-voice · P.P.P.S. = objection/placebo kill in-voice · final P.S. = reader-direct-address AS the CTA · then a **bare 👉 link on its own line AFTER the final coda, no CTA copy around it.** Note the link comes LAST, after the postscripts, not before them.
0228  - **The spoken-word reveal mechanic** — the product never enters as a decision. It enters as a word spoken by a named third party inside the story.
0229  - **The discovery chain** — article → forum thread → named stranger's comment → skeptical read → try. Narrating the chain does DR work: relatability plus implied social proof without making a claim.
0230  - **Category-then-brand split** — mid-story reveal names the category, the brand surfaces at the close.
0231  - **Priced failed-solution stacks** (70/95 of their primary copies), and their register hygiene: near-zero emoji, near-zero ALLCAPS, no comment-bait in the witness lane.
0232  - **Format-page discipline** — never cross formats across page identities.
0233  - **Unaware-pulling hooks:** pure mystery/curiosity openers that don't flag who the reader is. (This is the same failure the 8-31 SPNutrition analysis found from the conversion side: unqualified clicks. The awareness ruling and the targeting-clarity learning agree — hooks must name the reader by her observable symptom.)
0234  
0235  **STILL LIVE (unaffected):**
0236  - **T2 compressed witness** — problem-aware, and our ONLY scaled winner (fable-v10, $29.70 CPA). Never park this.
0237  - **T4 punch**, all Mode A work, objection killers, product-aware and offer-aware copy.
0238  - **T1 / Mode B when briefed to PROBLEM-AWARE traffic** — the architecture is legal, the unaware targeting is not.
0239  - The Layer 3 review check "is this written to the RIGHT stage?" stands and now has a narrower legal range: problem-aware → offer-aware.
0240  
0241  ## ⛔ LONG-FORMAT ARCHITECTURE = SOP 2.0 (Fish 2026-09-03, PERMANENT: "follow the doc and what it says. It is more true.")
0242  
0243  Every long-format native (anything over the T4 punch band) is written and validated against
0244  `references/long-format-native-sop-v2-contract.md`, the binding distillation of Zakaria's Long-Format
0245  Native Ad SOP 2.0. It wins over the Alevia-derived witness contract, the T1/T2 budgets, the
0246  victory-as-absence rule, the 55-70% reveal window, Mode B paragraph prose, and the analogy ban
0247  INSIDE the authority's mechanism dialogue. Load it before outlining. Structure in one line:
0248  hook (1-3) → the world (4-150) → inciting incident + authority (150-200) → mechanism through
0249  dialogue (200-400) → objections as story (400-500) → reveal at 75-80% by the authority (500-550)
0250  → week-by-week results ending on the relationship resolution (550-650) → close on loss, the honest
0251  line, guarantee, the "If by" ladder, 👉 link, the fork, thank-you, signature, P.S. ladder (650-700).
0252  Alevia paragraphs (2-4 sentences), quoted dialogue, clinician authority allowed, counted results, narrator
0253  identity block with her numbers (Fish 9-03 rulings after the measured Alevia comparison, see the contract's
0254  "Alevia deltas ruled in"). Fences that stay: no invented science, reader-side relatability gate, unaware
0255  lane parked, offer floor.
0256  
0257  ## Mode Selector — Mandatory
0258  
0259  Choose the format before writing. Do not blend architectures by accident.
0260  **Check the AWARENESS SCOPE GATE above before selecting any template.**
0261  
0262  ### STRUCTURE TEMPLATE — ALWAYS ASK FISH FIRST (binding, 2026-08-23)
0263  
0264  Four locked structure templates live in `references/native-structure-templates.md`
0265  (T1 Mode B Jacob-refined 3,200–4,000w · T2 compressed witness 800–1,100w · ~~T3 unaware
0266  study-reveal 550–700w~~ **[PARKED 8-31 — unaware lane]** · T4 punch 250–400w — each with per-section word budgets and
0267  reveal/CTA position markers). **Before writing ANY native, ask Fish which template to
0268  use** — one question, up front, per ad or per batch. Skip the question ONLY when the
0269  request or an outline handoff already names the template (the handoff format is in that
0270  reference). Fish's standing workflow: he picks the template and writes the section
0271  outline; the copywriter fills prose to the budgets and verifies word count, reveal %,
0272  and (T1) CTA position against the template before qa-gate. Never silently substitute a
0273  different structure than the one named.
0274  
0275  ### Mode A — Standard native execution
0276  
0277  Use the existing validated-brief, compressed Avelia T1/T2/T3, objection-killer, short, and mid-form workflows in this skill. This is the default only when neither the brief nor the supplied controls clearly selects long-form witness architecture.
0278  
0279  ### Mode B — DawnBands Long-Form Witness Native
0280  
0281  Use when the brief explicitly selects **DawnBands Long-Form Witness Native** mode **or when the user's supplied samples unambiguously specify that format**: roughly 2,500–4,000 words, first-person personal anecdote/witness story, a lived chronology before mechanism, and a late product reveal. “Like these samples” is a format instruction even if the user never types the internal mode name.
0282  
0283  Before choosing Mode A for a sample-led request, measure the controls' word counts, narrator type, and first product-name percentage. Never compress 3,000-word personal-story controls into 200–700-word ownership arguments merely because the request also asks for different personas, angles, or root causes.
0284  
0285  Load `references/control-format-selection-and-pilot-gate.md` for the control-measurement, narrator-selection, pilot, and back-half diversity gates.
0286  
0287  Before outlining or drafting, load `references/dawnbands-long-form-witness-native.md` and obey its narrator-authority, provenance-ledger, 2,500–4,000-word target, reveal-window, cue-vs-waking, outcome-timeline, recurring-motif, and DawnBands claim-ceiling gates.
0288  
0289  Do not treat a 500–700-word ownership ad as Mode B. Do not force Mode B onto product-aware traffic or compressed Avelia T1/T2/T3 briefs.
0290  
0291  ### Mode B first-draft execution (when intentionally selected)
0292  
0293  1. Load `references/dawnbands-long-form-witness-native.md`, `references/mode-b-first-draft-checklist.md`, and `references/control-format-selection-and-pilot-gate.md` before outlining. For measured live-account structure updates (paragraph-prose ruling, reveal distribution, spoken-word reveal mechanic, P.S.-ladder close variant, opener ceiling), also load `references/alevia-live-anatomy-2026-08-12.md`.
0294  2. Select narrator mode from the actual request and controls. When no verified Dawn buyer receipt exists, **ownership-voice** remains safest for argument-led work; however, when Fish explicitly asks for personal anecdotes or supplies first-person witness controls, use internally labeled **authorized-dramatization** rather than flattening the format into detached ownership prose. Keep behaviors research-grounded and never invent DawnBands outcomes, medical authority, quantified proof, or a named fake testimonial.
0295  3. Fill the internal YAML contract first: submode, narrator mode, motif, primary + ≥1 secondary open loop, reveal target, exact-SKU claim classes, offer_policy_verified, product_outcome_receipt null/ID.
0296  4. Draft to file under `output/.../copy/` with **Fable authoring in-session as the default copywriter** (amended 2026-08-12 — the only scaled winner, fable-v10, was a Fable draft; Grok/another model only when Fish explicitly names one). Preserve the exact first draft as a separate immutable artifact before review. Count primary-text words and first product-name % after every expand pass. Mode B floor is **2,500 words** — do not ship under-floor drafts labeled Mode B.
0297  5. For multi-ad Mode B batches, draft and audit one complete pilot before scaling. Read the opening, reveal window, and close; a good hook does not excuse a detached or cloned back half. After scaling, compare product bridges, mechanism lists, offer cadence, CTA scaffolds, and motif returns across the whole batch. If they converge, preserve the fronts and run same-model surgical back-half rewrites.
0298  6. Offer block: verify current price + refund terms. Prefer unambiguous **60-night money-back** language. Omit **100% wake-up guarantee** unless legal/substantiation approved it as a named refund policy and the line is clarified as refund terms, not universal waking efficacy.
0299  7. Victory = desired absence / test frame unless a same-speaker product receipt exists. No invented week-by-week results.
0300  8. After the immutable first draft, run the **Surgical Review Pass** (section 7): one line-edit pass against the four layers + format + hook criteria. GPT roast / Sonnet comparison artifact only on the section-7 escalation triggers (new-lane pilot, Fish request, structural failure). The Opus taste pass is retired. Fish’s explicit model request wins.
0301  9. For artifact freezing, hashes, measured QA, narrator-provenance review, provider retry, and the three-column Cloudflare comparison report, load `references/model-routed-copy-review.md`.
0302  10. Worked ownership-voice Mode B example: `/Users/ayden/.openclaw/workspace/output/dawnbands-native-2026-07-19/copy/ad-21-door-memory.md` (architecture reference only — do not clone lines; its routed review found pooled-VOC narrator risk, so it is not a spend-ready testimonial model).
0303  11. **Reading deliverable:** When Fish wants to read a finished Mode B / long native, ship a **public Cloudflare Pages URL** (not only a workspace path). Recipe in `references/mode-b-first-draft-checklist.md` → Delivery. Dedicated Pages project per report — never clobber an unrelated batch field-report project.
0304  
0305  ## Native-First Priority (Fish, 2026-07-19)
0306  
0307  For ad prose, optimize in this order:
0308  
0309  1. **Sounds native** — like a real post or lived thought, not a copywriter performing authenticity.
0310  2. **Story architecture is coherent** — rupture, escalation, discovery, mechanism, reveal, and payoff form one causal chain; each beat makes the next beat necessary.
0311  3. **The mechanism makes sense and sells** — for DawnBands, use the canonical Deep-Sleep Signal Lock biological stack, explain why repeated sound failed, then make the tactile bypass and Wrist-First Handoff feel inevitable. Do not neuter it into feature-only “different channel” language.
0312  4. **Breakthrough Advertising strategy is active** — channel existing mass desire, match awareness and sophistication, complete the belief chain, externalize the villain, and use the mechanism/proof depth the market requires.
0313  5. **Product arrives naturally** — the reader should understand the needed solution before the product appears; never announce the sales pivot.
0314  6. **Conversion language stays invisible** — preserve useful roughness, ordinary phrasing, and human irregularity. A more polished sentence is a regression when it sounds more like an ad.
0315  7. **Coherence beats cleverness** — remove quote-card abstractions, theatrical metaphors, internal QA language, and obvious objection-module formatting.
0316  
0317  Sonnet polish is not a beautification pass. It may tighten drag and repair flow, but it must not sand the copy into perfect sales language or weaken the story, mechanism, awareness match, sophistication match, belief chain, or proof logic. Final read tests:
0318  
0319  - **Would this sound natural if the target avatar posted it without knowing she was writing an ad?**
0320  - **Does every story beat advance the belief chain?**
0321  - **Does the mechanism make prior failure and the product's bounded role feel logically inevitable?**
0322  
0323  ## Existing Native Review — Full Context Intake
0324  
0325  When reviewing, revising, or explaining an existing DawnBands native, do not stop at the public/rendered copy. Verify and read the source stack that produced it: current skill rulings, format contract, current avatar/sub-avatar work, structured unit brief, prompts, brand truth, VOC receipts, final copy, and its human + deterministic audits. Keep current governing doctrine separate from historical examples and archived research.
0326  
0327  For the exact intake order, Evolve source map, inciting-incident precedence, and minimum existing-batch artifact set, read `references/dawnbands-context-intake-and-precedence.md`.
0328  
0329  ## Canonical Source Sync
0330  
0331  This Hermes skill has a Claude Code counterpart at `/Users/ayden/.openclaw/workspace/.claude/skills/native-copy/SKILL.md`.
0332  
0333  Cadence amended 2026-08-12: run this comparison WEEKLY or when output looks off-doctrine — not
0334  before every run (it was a per-run process tax with near-zero catch rate).
0335  1. Compare the Hermes and Claude Code versions for drift.
0336  2. If the Hermes skill is missing or incomplete, consult the Claude Code version before writing.
0337  3. If they differ, use the newest writing doctrine from the Claude Code version while preserving Hermes-specific safety and tool instructions from this skill. Never silently rely on a stale imported copy.
0338  4. When checking adjacent integration configs, never print credential values. Report only presence, validity, and source path.
0339  
0340  ## Workflow
0341  
0342  ### 0. Read Competitive Landscape (MANDATORY)
0343  Read `knowledge/research/competitive-landscape.md` — contains saturated angles (HIGH CPM, avoid), white space angles (LOW CPM, prioritize), competitor hooks to steal/counter, and positioning rules. This step is NOT optional.
0344  
0345  If a doctrine path has moved, search the workspace for the filename before concluding it is missing. Once the canonical location is verified, update this skill rather than silently skipping the gate.
0346  
0347  **Before writing ANY copy, verify:**
0348  - The angle you're writing is NOT in the "Saturated Angles" list (or has a clear differentiation plan)
0349  - You're prioritizing white space angles (60%+ of new batch)
0350  - You're not duplicating a competitor's exact hook or framing
0351  - You're incorporating at least one stolen copy device (adapted to our positioning)
0352  
0353  ❌ NEVER write another "I shake my ADHD child awake" ad — that's Nymera's saturated lane
0354  ❌ NEVER write all-fear copy with zero aspiration — every competitor does that, we differentiate with hope
0355  ❌ NEVER skip this step. Copy written without competitive context = wasted tokens
0356  
0357  ### 1. Read the Brief
0358  Read `references/master-brief.md` — contains everything: avatar, mechanism, format rules, Zak framework, dead angles, copy killers, headline rules, targeting rules.
0359  
0360  ### 2. Check the Gold Standard
0361  Read `assets/gold-standard-example.md` — correctly formatted variants with headlines. Match this format exactly.
0362  
0363  ### 3. STANDING NATIVE TEST LAYOUT (Fish, 2026-08-23 — supersedes the same-day 3×3 grid and the old 5-variant default)
0364  Every native test unit ships as:
0365  - **1 body script** — one desire × one sub-avatar × one message. Length set by what the angle needs (7-22 no-caps ruling); template still ALWAYS asked (Mode Selector).
0366  - **3 hook variations** — line 1 swapped on the same body, each graded on the hook taxonomy (structure named, curiosity/validation mixed). Hooks are the copy-side test lever.
0367  - **5 image concepts** — the primary creative lever (receipt: image swap moved fable-v10 + its sniper test hard). Scrape-first per native-image-gen; mix concepts across the operator families: contradiction pair (two objects that don't belong together, zero copy on image), problem-proof "help-post" shot, artifact/rupture object, straight rip. No product, no text overlay, phone-photo register.
0368  - **2+ headlines authored — ALL load as title options at launch (Fish 2026-08-25 + 2026-08-29, supersedes the 8-23 "picks ONE, never split-tested" rule).** Meta standard ads take up to 5 titles. Every hook and every headline rides in at launch via the solved multi-text construction. Nothing waits for a pick.
0369  - **Length variants are a legal test axis inside one unit (Fish 2026-08-31).** A unit is one ARCHITECTURE (one desire × one sub-avatar × one message); long/mid/short cuts of that architecture ride as body slots in the multi-text stack (Meta caps at 5 bodies — spread them so every length and every hook appears at least once). This does not license two different angles in one unit.
0370  - **Validated structured brief:** the brief's `goal.variant_count` / `output_contract` still override when explicit.
0371  
0372  ### 3a. OFFER FLOOR — every native ships with a complete close (Fish 2026-08-31, BINDING)
0373  
0374  Receipt: `v11-a-unassigned-hour` drew a 4.13% link CTR at $0.92 CPLC with NO brand name, NO offer, and NO link anywhere in the body. Clicks were manufactured by withholding every resolution; CPA landed $46-56. Every ad that actually converts in the account (fable-v10, r2-d2-roommate, slop 7, the offer-ad family) carries a complete close.
0375  
0376  **REQUIRED in every native body, every length, every destination:**
0377  1. **Brand named at the reveal** — "Dawn Bands", two words. A body that never names the brand is not a native, it is a click generator.
0378  2. **The guarantee, stated plainly** — "100% wake-up guarantee. 60 nights." Say what it means in her language at least once ("60 nights to let it prove it on school mornings"). The guarantee, not the price, is the mandatory risk-reversal element.
0379  3. **Company/credibility intro, 2-4 lines — and it MUST NAME THE BRAND as its subject** (Fish 2026-08-31, second ruling). Write "Dawn Bands is a small company and this is the only thing they make." NEVER "The company that makes it is small" or "The company is small" — an anonymous company intro is a floor FAILURE, not a stylistic choice. Receipt: in the 8-31 model bake-off, 3 of 5 drafts satisfied "brand at reveal" and "company intro" as separate items and produced an orphaned, brandless credibility block. The brand name has to be the subject of that sentence. Small-operator framing beats corporate framing; carry the quality reason into a natural availability consequence.
0380  4. **At least one link.**
0381  5. **P.S. ladder + TWO CTAs — allowed and encouraged on long copies** (Fish 2026-08-31). Close order is now fixed:
0382  
0383     `body → CTA #1 with link → P.S. ladder → CTA #2 with link`
0384  
0385     **One CTA before the postscripts and one after** (Fish 8-31, third ruling). This amends the Alevia measured default of link-last-only — their ladder ends on a single bare link; ours brackets the ladder so a reader who stops at the first postscript still has a live link above them and a reader who reads to the end gets another. Rungs, per the Alevia shape: P.S. = rupture-object return · P.P.S. = guarantee naturalized in-voice · P.P.P.S. = objection kill in-voice · final P.S. = reader-direct-address. Optional on short/punch formats.
0386  
0387  **OPTIONAL — use when it earns its place:** price ($49 may be woven in or omitted; Fish 8-31: the offer and guarantee are required, the number is not), sale/availability framing, the narrator's own purchase beat, P.S. ladder.
0388  
0389  **DESTINATION:** listicle and PDP both convert. The floor is identical for both. A listicle ad must be FULLY CONGRUENT with the page it lands on — same claim ceiling, same mechanism, same offer terms. Incongruence kills listicle traffic, not the destination type itself.
0390  
0391  **Variation lives in HOW these are woven into the story, never in WHETHER they exist.** This supersedes master-brief's CTA-variation permission "No mid-copy CTA at all — just the close CTA" where it would leave a body with no close. The price-omission permission survives; the missing-close permission does not.
0392  
0393  #### Reference close-block shape (SPNutrition-measured; adapt the register, never the lines)
0394  1. Company intro: small company, limited batches, the quality mechanism restated in plain words, contrasted against the cheap category default.
0395  2. Quality → availability consequence: that level of care takes time, which is why restocks take weeks.
0396  3. Current availability plus the narrator's own action ("I ordered a second one").
0397  4. Guarantee in plain language: the window, what happens if nothing changes, no questions asked.
0398  5. Conditional future-pace ladder: "But if it works? If by [early marker]... If a month from now [the desired absence]?" Three escalating conditionals, each a desired ABSENCE, never a promised outcome.
0399  6. Link.
0400  7. Closing callback returning to the hook thesis in one line.
0401  
0402  Claim gates unchanged: the ladder stays conditional and absence-shaped so it never becomes an invented outcome timeline. Never invent batch sizes, customer counts, stock levels, or sale end dates. Availability language must be true at launch or omitted.
0403  
0404  #### Validated Brief Execution Mode
0405  When the user supplies a validated brief batch (for example `briefs.json`), treat each brief as an executable contract:
0406  1. Extract only the requested brief IDs rather than bulk-loading the whole batch into context.
0407  2. Follow the brief's exact word range, fields, destination type/URL, CTA, awareness, mechanism-delivery mode, and output path.
0408  3. The newest live evidence source governs claims: an exact-SKU/product-mechanism report or current policy receipt overrides an older brief, canonical product block, gold standard, or brand-doctrine example. If sources conflict, stop the affected claim, record the contradiction, and use the narrower live claim ceiling rather than silently inheriting the older language.
0409  4. Preserve exact offer wording/order only after verifying that the current policy and newest live claim report authorize both the terms and the consumer-facing implication. For DawnBands, do not mechanically carry `100% wake-up guarantee. 60 nights. $49.` from an older brief when a newer report treats guaranteed waking or exact-SKU efficacy as unsubstantiated. If the policy is verified but the name could imply universal efficacy, distinguish the refund-policy terms from an outcome claim and flag for legal/substantiation review.
0410  5. Include a compact declaration block and source VOC IDs when required so the artifact is auditable.
0411  6. Do not invent customer outcomes, timelines, reviews, or first-person testimonials to satisfy a story format. Use supplied VOC as attributed evidence, or write a clearly non-testimonial POV/ownership argument that does not claim an unverified result. If no product outcome is evidenced, express the victory as the avatar's desired absence rather than claiming it happened.
0412  7. Even when a brief labels the entry door a mechanism reveal, line 1 must still open on a lived artifact, contradiction, or emotional tension. Reveal product/mechanism only after the hook.
0413  8. Write only the authorized deliverables. If the scope ceiling names ad files plus one manifest, do not add strategy reports, research notes, or extra artifacts.
0414  9. Run deterministic validation ONCE after the surgical review pass (and again only if a later edit touches the text): actual primary-text word count, required fields, offer order normalized across line breaks, destination, banned claims/price, em dashes, line format per the mode-scoped rule in the Format Validation Pass (sentence-per-line for Mode A; paragraph prose for Mode B), dialogue isolation, declared VOC IDs. SHA-256 manifest hashes are required only when the validated brief's contract demands a manifest — not on ordinary batches (cadence amended 2026-08-12; the after-every-edit loop was pure process tax).
0415  
0416  Reusable execution detail lives in:
0417  - `references/lived-first-person-native-copy.md` for repairing detached analyst voice into lived first-person native copy without inventing a customer outcome.
0418  - `references/validated-brief-execution.md` for authority order, truth-safe POV writing, and deterministic per-ad QA.
0419  - `references/validated-brief-batch-contract.md` for authority order, scope discipline, claim-safe POV handling, canonical-offer line breaks, and deterministic manifest QA.
0420  - `references/brief-driven-batch-validation.md` for the existing brief-batch validation checklist.
0421  - `references/native-proof-and-architecture-overlays.md` for testimonial provenance, long-form-vs-compressed architecture, live-Meta evidence use, cross-ad convergence, and the repair/re-QA sequence.
0422  - `references/long-form-native-comparative-audit.md` for read-only batch audits against winner copies, a live long-form SOP, and live avatar/mechanism reports; it defines source precedence, deterministic reveal measurements, device/per-ad matrices, structural-vs-provenance separation, and brand-doctrine output.
0423  - `references/consolidated-native-ad-repair-pass.md` for repairing a requested ad subset from multiple QA reports, truth-safe conversions, blocker-specific assertions, and final manifest/hash revalidation.
0424  - `references/controlled-live-meta-repair.md` for repairing a brief-driven batch under a strict file ceiling while adding a small number of controlled live-evidence mechanism lanes without cloning winner copy.
0425  
0426  ### Editorial truth gate before paid creative generation
0427  
0428  Deterministic PASS is necessary but not sufficient. Before image prompts or paid generation:
0429  1. Run a flags-only editorial/source-fidelity audit across the whole batch.
0430  2. Treat synthesized first-person scenes, unsupported precision, altered quotations, invented dialogue, and composite timelines as blockers even when every underlying behavior is VOC-backed.
0431  3. Convert blockers to attributed VOC, explicit hypotheticals, paraphrases without quotation marks, ownership arguments, or desired-absence framing.
0432  4. Run a separate cross-ad convergence check for mechanism sentences, product descriptions, narrator cadence, analogies, victory images, and CTA scaffolds.
0433  5. If a new SOP or live performance export appears while QA is running, add one overlay audit and then make a single consolidated repair pass rather than rewriting the batch repeatedly.
0434  6. Consolidate repairs into ONE pass, then re-run deterministic validation and editorial QA once on the repaired batch (amended 2026-08-12 — not after every individual edit).
0435  7. Keep image generation gated until the repaired copy passes.
0436  
0437  ### 3.5 Multi-Angle Batch Architecture
0438  
0439  When Fish asks for many ads with different personas, angles, stories, root causes, or mechanism routes, do not rely on a prose instruction to “make them different.” Build a row-per-ad diversity matrix first, then route bounded model lanes from that shared matrix.
0440  
0441  Load `references/multi-angle-batch-routing-and-repair.md` for:
0442  - required matrix dimensions
0443  - dominant-root-cause distribution
0444  - raw-versus-final artifact structure
0445  - deterministic and editorial convergence checks
0446  - same-model surgical repair when a full agent prompt is too expensive
0447  - public batch report delivery
0448  
0449  Truth and provenance rules stay in the hidden brief. **Never let them leak into consumer copy** as lines about not inventing reviews, not promising outcomes, avoiding fake stories, or clarifying internal claim policy. Replace that leakage with natural skepticism, a bounded test, risk reversal, or avatar-native objection language.
0450  
0451  ### 4. Write Variants
0452  Each variant uses a different angle from the 4-type framework:
0453  1. **Story/Resignation Arc** — first person narrative, resignation moment, discovery
0454  2. **Objection Killer** ← TOP PERFORMER (Mar 2026). Write 2 per ad when possible. "Another alarm? Sure." energy converts.
0455  3. **Pain-First → Mechanism Reveal** — vivid daily pain, mechanism in the middle
0456  4. **Consequence/Future Fear** — what happens if nothing changes
0457  5. **(If needed) Risk Reversal + Social Proof** — reviews + guarantee forward
0458  
0459  **Mode A length (Fish 2026-08-23 test layout):** ONE body per test unit; pick its length by how much DR work the angle needs (150-300w short, 500-700w mid, 800-2,100w long band — no caps). Variation budget goes to hooks (3) and images (5), not parallel bodies. **Mode B** uses `references/dawnbands-long-form-witness-native.md` (2,500–4,000 words) and is never satisfied by a 500–700-word draft. See also `references/mode-b-first-draft-checklist.md`.
0460  
0461  Each variant needs:
0462  - **Headline** (Meta `title` field)
0463  - **Body copy** (Meta `body` field)
0464  - **Destination URL** (listicle, not PDP)
0465  
0466  ### 5. Format Validation Pass (MANDATORY)
0467  After writing ALL copy, do a final scan.
0468  
0469  **Line format is MODE-SCOPED (corrected 2026-08-12 per the 7-31 Mode B ruling):**
0470  - **Mode A / compressed natives / static primaries:** every line = 1 sentence (2 max if flow
0471    demands it); NEVER 3+ sentences on one line.
0472  - **Mode B / long-form witness:** PARAGRAPH prose — 2–4 sentences per paragraph (~25–40 words),
0473    blank line between paragraphs; staccato single-lines only at punch moments. One-sentence-per-
0474    line formatting in Mode B is a format defect, not a style choice
0475    (`references/alevia-live-anatomy-2026-08-12.md`).
0476  
0477  All modes:
0478  - No em-dashes
0479  - No emoji except 👉 on CTA links (and the T4 checklist template, where ✅ is the register)
0480  - Dialogue isolated on its own line
0481  - No dead angle hooks (mechanism-first, generic features, curiosity bait)
0482  - No copy killers ("game changer", "literally", "amazing", vague future pacing)
0483  
0484  ### 6. Hook Sharpening Pass (MANDATORY)
0485  After writing all variants, re-read every hook (line 1 of each variant's body copy). Grade each:
0486  - **WTF?** Does it stop scroll? Does it create a question she HAS to answer? Would she screenshot it and send it to a friend?
0487  - **Specific?** Does it contain a concrete detail (number, name, object, timestamp)? Generic hooks die.
0488  - **Gap?** Does it open a loop that can only be closed by reading more?
0489  - **Taxonomy (Fish 2026-08-23, `~/.openclaw/workspace/knowledge/hook-taxonomy.md`):** name the emotion it fires, the stake on the table, and the structure — **Promise** (bold claim upfront, exceptionally clear) or **Loop** (key detail withheld). "Both/neither" = rewrite. Execution screen: zero backstory required, starts in action, no clutter — when in doubt, dumb it down.
0490  - **Targeting clarity (Fish 2026-08-31, the 5th grading question):** could a stranger tell from lines 1-2 whether this ad is about HER house? Name the observable symptom or daily job it flags. "It builds intrigue" is not an answer. A Loop hook is only legal when the withheld information is about the READER'S OWN situation, never generic mystery. Receipts: native1 24.8% CTR / 0 purchases; v10 body $8,267 at $51 CPA opening on a victory brag that reaches identification nine lines down; slop 7 flags her in line 1 and closes at $29 CPA on $14,475.
0491  
0492  **If a hook fails any of these, rewrite it before submitting.**
0493  
0494  Common hook failures to catch:
0495  - ❌ Too bare: "He set it himself." → needs the MOMENT: "I walked past his room. He was adjusting something on his wrist. He didn't know I was there."
0496  - ❌ Too generic: "I used to be the kid everyone had to wake up." → needs the SPECIFIC story: "I slept through a fire alarm at a sleepover. Everyone thought something was wrong with me."
0497  - ❌ Too subtle: "I just didn't go in." → needs the CONSEQUENCE: "One morning I didn't go in. By 11:40 the school called me."
0498  - ❌ Statement without tension: "His teacher emailed me last week." → needs the CONTENT: "His teacher emailed me two words: 'perfect attendance.' I stared at my phone for ten minutes."
0499  - ❌ Product listing as hook: "Sonic Boom bed shaker. $45." → needs the TOTAL + tease: "$380. That's what I spent before someone told me they all use the same pathway his brain blocks."
0500  
0501  **The hook is the only line that matters at scale. If it doesn't stop scroll, the rest is irrelevant. Spend more time on hooks than on any other part of the copy.**
0502  
0503  ### 7. Surgical Review Pass (MANDATORY — one pass, line edits only)
0504  
0505  **(Restructured 2026-08-12, Fish-approved lean pipeline.)** The old mandatory multi-model
0506  stack (separate Sonnet full-polish artifact + GPT roast + optional Opus) ran up to seven
0507  text-mutating passes per ad. The only scaled winner (fable-v10, $5.4k @ 2.28 lifetime) never
0508  went through it, and this skill's own Native-First rule says a more polished sentence is a
0509  regression when it sounds more like an ad. Rewrite passes sand off the native roughness that
0510  IS the format.
0511  
0512  The default is now ONE review pass:
0513  
0514  - Save the exact first draft unchanged (immutable, one file copy — keep this).
0515  - The drafting model re-reads the draft against the **four layers below** plus the Format
0516    Validation Pass and the Hook Sharpening criteria, and fixes findings as **line edits** —
0517    never a top-to-bottom rewrite. Blockers (claim/provenance/offer/format) MUST be fixed;
0518    taste flags are logged for Fish, not auto-applied.
0519  - Output: the repaired artifact + a short blocker/change log. Not three artifacts.
0520  
0521  **Conditional escalation** (any of these, not by default): a NEW angle or lane's pilot ad,
0522  Fish asks for it, or the surgical pass finds structural failure (rebrief-level problems):
0523  - **GPT challenger roast** — decision memo on original + repaired, never a silent rewrite.
0524  - **Sonnet full-polish artifact** — only when Fish explicitly wants a comparison version.
0525  
0526  **Humanizer (amended 2026-08-12 with Fish's approval — supersedes the 2026-07-20 always-on
0527  rule):** run the humanizer pass only when (a) the draft came from a model other than Fable,
0528  or (b) the repaired draft fails a native-read spot check — read the opening 8 lines and one
0529  mid-story passage cold: would this pass as a person posting? Fable in-session drafts that
0530  pass the spot check ship without the humanizer. All DR beats are preserved whenever it runs.
0531  
0532  The review layers (the checklist for the surgical pass):
0533  
0534  **Layer 1: Hook Audit**
0535  - WTF factor (1-10). Anything below 7 = flagged with rewrite suggestion
0536  - Is it something she's heard before? Check against hook bank dead list
0537  - Does it open a gap only the copy can close?
0538  - ❌ NEVER pass a hook that reads like a headline. Hooks = native voice, not ad voice.
0539  
0540  **Layer 2: Copy Quality**
0541  - Flag AI-sounding language (graveyard, journey, struggle, battle, warrior, game-changer, literally)
0542  - Flag weak/filler lines that don't earn their space
0543  - Flag soft CTAs or templated CTA blocks
0544  - Flag pacing issues (too many long sentences in a row, no breathing room)
0545  - ❌ NEVER just say "this line is weak." Provide the replacement line.
0546  
0547  **Layer 3: Framework Alignment**
0548  - Awareness level: is this written to the RIGHT stage? (unaware ≠ solution-aware)
0549  - Mechanism: is it specific enough? Does it feel like the first time she's heard it?
0550  - Resignation moment: is it present? Is it EARNED (not just stated)?
0551  - Villain: is there a clear external force she can blame?
0552  - Desire magnitude: is the pain/want big enough to act on?
0553  - ❌ NEVER pass a variant missing both resignation moment AND villain. At least one must be present.
0554  
0555  **Layer 4: Avatar Fidelity**
0556  - Does this sound like HER words or a copywriter's words?
0557  - Cross-check against VOC word bank in avatar interior doc
0558  - Check the 3am thoughts, the parking lot moment, the "I've heard that before" exhaustion
0559  - Is the victory written as what STOPPED (not what started)?
0560  - ❌ NEVER pass copy where the avatar sounds aspirational instead of exhausted. She's resigned, not motivated.
0561  - **Lines 3-8 specificity test:** Read lines 3-8 of body copy. Broad/general = avatar was assumed, not found in VOC. Specific/concrete = avatar was found. "Many women feel frustrated when their dental routine doesn't deliver results" = assumed. "The hygienist said excellent. Again. For the third year in a row. While her teeth were getting worse." = found. Flag any copy that fails this test — rebrief required, not a line-level fix.
0562  
0563  **Synthesis rules (surgical pass + any conditional escalation):**
0564  - Truth/provenance/offer/format breach → mandatory line-edit fix before the artifact can pass
0565  - Taste/conversion disagreement → logged for Ayden's judgment, never auto-applied
0566  - Cadence/specificity improvements → only as line edits that preserve provenance, claim
0567    ceiling, and the draft's native roughness
0568  - Never mutate the immutable first draft; repairs land in the working artifact
0569  - Side-by-side comparison surfaces (original / repaired / roast memo) are built ON REQUEST,
0570    not by default
0571  
0572  ### 7.5 Independent Adversarial Audit Mode
0573  When reviewing a completed ad batch rather than writing it, do not treat deterministic validation as editorial approval.
0574  
0575  Run two explicit gates:
0576  1. **Artifact contract:** files, fields, word ranges, offer order, destinations, and formatting.
0577  2. **Editorial/provenance:** hook quality, lines 3–8 specificity, avatar fidelity, claim/testimonial truth, awareness, persuasion chain, and cross-ad diversity.
0578  
0579  A deterministic PASS can coexist with an editorial FAIL. In particular:
0580  - A brief-authorized scene is not evidence that the scene happened.
0581  - VOC that supports a behavior does not authorize inventing a first-person narrator, chronology, dialogue, outcome, or precise number.
0582  - Plausible precision is still unsupported precision.
0583  - Repair composites by attributing exact VOC, using explicit hypothetical framing, writing a non-testimonial ownership argument, or deleting the unsupported detail.
0584  - In read-only audit mode, provide exact replacement proposals for every score below 7 without editing source ads.
0585  - Separate high-confidence blockers from subjective taste flags and label confidence.
0586  
0587  Use `references/adversarial-batch-copy-audit.md` for the scoring matrix, provenance test, cross-ad diversity checks, and deliverable verification. When the audit compares a batch with competitor/winner copies, a live long-form SOP, and live avatar/mechanism reports, also use `references/long-form-native-comparative-audit.md`; it requires an explicit source-precedence ledger, measured reveal timing, a winner/SOP device matrix, a per-ad disposition ledger, and separate structural versus truth/provenance findings.
0588  
0589  ### 8. Cross-Variant Diversity Check
0590  - Each variant must hit a genuinely different Persona × Desire × Awareness combo
0591  - No two variants should open with the same hook formula
0592  - Headlines must not duplicate across variants
0593  - No testimonial names shared with listicle comments
0594  - **Mechanism delivery MUST be different in each variant** (full science / casual / secondhand / barely explained / analogy / skipped — see master-brief)
0595  - **CTA format MUST vary** — do NOT use "👉 Here's [reason]:" on every variant
0596  - **At least 1 variant must be SHORT** (15-20 lines) — not every variant is a 40-line story
0597  - **Social proof reviews must sound real** — typos, incomplete thoughts, 4-star mixed in, platform references
0598  
0599  ## ⛔ NEVER DO (check EVERY time)
0600  - [ ] 3+ sentences on one line → KILL IT
0601  - [ ] Em-dashes anywhere → KILL IT
0602  - [ ] Emoji on CTA that isn't 👉 → KILL IT
0603  - [ ] Hook that matches anything in dead angles list → KILL IT
0604  - [ ] "Game changer", "literally", "amazing", vague future pacing → KILL IT
0605  - [ ] Mechanism-first opening (leading with how the product works) → KILL IT
0606  - [ ] Same CTA format on every variant → FIX IT
0607  - [ ] All variants are 35+ lines (no short one) → FIX IT
0608  - [ ] Testimonial names that match listicle comment names → FIX IT
0609  - [ ] Didn't read master-brief.md before writing → STOP AND READ IT
0610  - [ ] Hook is scene-setting or chronological start instead of most intriguing element → REWRITE HOOK
0611  - [ ] Best line is in headline field instead of primary text line 1 → MOVE IT
0612  - [ ] Same mechanism paragraph across multiple variants → REWRITE EACH DIFFERENTLY
0613  - [ ] Same analogy/metaphor repeated across variants → KILL DUPLICATES
0614  - [ ] AI-default words (graveyard, journey, struggle, battle, warrior) → REPLACE WITH UGLY HUMAN WORDS
0615  - [ ] Internal QA/provenance leakage ("I won't invent a review," "no fake before-and-after," legalistic claim caveats) → REWRITE AS NATIVE SKEPTICISM OR A BOUNDED TEST
0616  - [ ] CTA block is templated/identical across variants → REWRITE TO FLOW WITH STORY
0617  - [ ] Victory written as what started ("mornings are peaceful") instead of what stopped ("I didn't yell") → REWRITE
0618  - [ ] All 5 variants same persona + desire + awareness level → DIVERSIFY AT LEAST 1 DIMENSION
0619  - [ ] Mode A: dense paragraphs instead of 1 sentence per line with breathing room → BREAK APART (Mode B is PARAGRAPH prose by contract — do NOT break apart a witness native)
0620  - [ ] ⛔ ALWAYS check the full ⛔ NEVER DO section in master-brief.md — it is the single most important section
0621  
0622  ## Avatar Detail Extraction (MANDATORY — do this before writing copy for ANY avatar)
0623  
0624  Before writing copy, extract AT LEAST these categories from the avatar research doc into the subagent brief. If any category is missing from the research, flag it — don't skip it. If the research reveals additional categories not listed here, include those too. These are the details that separate copy that converts from copy that reads like AI.
0625  
0626  1. **Failed solutions stack WITH visceral detail.** Not just "tried alarms." The specific unsettling behavior: he walks to it in his sleep, eyes open, solves math puzzles unconscious. The detail that makes a reader go "wait what?"
0627  2. **Anticipation anxiety.** What happens BEFORE the problem starts each day. She wakes before her alarm from dread. The cortisol spike. The mental math of how much buffer she has.
0628  3. **The scene after.** The drive to work in silence. The parking lot. The replaying. The 10 minutes before she can go inside. Concrete, visual, usable in copy.
0629  4. **The "I've heard that before" trap.** Everyone thinks they have the answer. She's exhausted by suggestions. This is the skepticism objection killers need to mirror.
0630  5. **Relationship cost.** What the problem did to the relationship over time. Not one fight — accumulation. He stopped talking. She became someone he avoids. The repair moment when it starts working.
0631  6. **Victory as ABSENCE.** What STOPPED happening. Not what started. She hears footsteps before she went in there. She didn't yell today. Write victory as disappearance, not arrival.
0632  7. **VOC word bank.** Her actual words organized by emotional state. Default to these over AI-generated language. Always.
0633  
0634  If a subagent doesn't have all 7, the copy will be generic. Every time.
0635  
0636  ## Subagent Handoff Rule
0637  When spawning a copy subagent, it MUST receive:
0638  1. The full `references/master-brief.md` file (not a summary)
0639  2. The `assets/gold-standard-example.md` file
0640  3. The specific ad copy it's writing variants for
0641  4. The NEVER DO checklist above (copy-paste it into the task)
0642  5. The 7 avatar detail categories above — extracted from avatar research, not summarized
0643  Do NOT summarize the rules. Pass the files directly.
0644  
0645  ## Key Files
0646  - `references/native-structure-templates.md` — **the four locked structure templates (T1–T4) with per-section word budgets; template choice is ALWAYS asked of Fish before writing (see Mode Selector)**; Fish-outline → copywriter-fill handoff contract lives here
0647  - `references/matrix-angle-execution-map.md` — **load FIRST when the brief starts from the ad-batch-hub testing matrix** ("expand my tests", "feed cell X"): resolves every matrix angle column (incl. wrong-label / defiance / witness, added 8-12) into definition, Dawn form, mode/length, receipts, and claim gates; also the proof-column devices (comparative two-person, survey-voice, retro-proof)
0648  - `references/control-format-selection-and-pilot-gate.md` — infer long-form mode from supplied controls, select personal-anecdote narrator authority, pilot one complete ad before scaling, and repair cloned reveal/close scaffolds without rewriting strong story fronts
0649  - `references/multi-angle-batch-routing-and-repair.md` — class-level workflow for pre-draft diversity matrices, bounded same-model lanes, immutable raw artifacts, convergence QA, provenance-leak cleanup, exact-model micro-repairs, and public batch reports
0650  - `references/dawnbands-selling-root-cause-mechanism.md` — compact execution reference for the canonical Deep-Sleep Signal Lock → Somatic Wake Bypass™ → Wrist-First Handoff™ selling doctrine; load for every DawnBands format so proof review does not flatten the mechanism into generic cue-placement language
0651  - `references/dawnbands-long-form-witness-native.md` — canonical Avelia/Manus-derived DawnBands witness-story mode: narrator authority, provenance ledger, 2,500–4,000-word architecture, bounded mechanism, motif return, outcome gate, and final QA
0652  - `references/mode-b-first-draft-checklist.md` — Mode B operator card: YAML gates, offer/SKU traps, under-floor wordcount, worked example, Cloudflare Pages reading-report deploy
0653  - `references/master-brief.md` — full copywriting brief (avatar, mechanism, format rules, dead angles)
0654  - `references/adversarial-batch-copy-audit.md` — read-only batch audit protocol for scoring, testimonial provenance, unsupported precision, blocker/taste separation, and diversity verification
0655  - `references/native-proof-and-architecture-overlays.md` — testimonial truth, long-form Manus architecture boundaries, live-Meta evidence use, convergence checks, and the pre-image repair sequence
0656  - `references/avelia-short-native-templates.md` — verified competitor-derived T1/T2/T3 structures for compressed short native copy; use as a structural reference, never as winner proof
0657  - `assets/gold-standard-example.md` — cadence/format example only; do not inherit its product truth, claims, scenes, dialogue, timelines, or outcomes without independent receipts

Native copy / assets/gold-standard-example-old-adhd.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/assets/gold-standard-example-old-adhd.md

SHA-256: db92e7f772c9bdba1f17462128b6ba0ff164c74878163354bd9dcb691aa1501b

Full source · 292 lines
0001  # Native8 — ADHD Listicle — 5 Copy Variants (Gold Standard)
0002  **Image:** native8 (ADHD med cycle native)
0003  **Destination:** dawnbands.com/pages/adhd-teen-alarm
0004  **Date:** 2026-03-05
0005  
0006  ---
0007  
0008  ## VARIANT 1 — Native/Story (Resignation Arc)
0009  **Awareness:** Level 2 (Problem Aware) | **Persona:** ADHD Mom | **Emotion:** Exhaustion → Relief
0010  
0011  **Headline:** I Did Everything His Therapist Told Me. He Still Couldn't Wake Up.
0012  
0013  I set up the perfect ADHD bedtime routine.
0014  
0015  Same time every night.
0016  Phone out of the room.
0017  White noise machine.
0018  Melatonin at 8:30.
0019  Lights off at 9:15.
0020  
0021  I followed every protocol his therapist gave me.
0022  
0023  He still couldn't wake up.
0024  
0025  Not "didn't want to."
0026  Couldn't.
0027  
0028  I've watched him open his eyes, say "I'm up," and be completely unconscious.
0029  
0030  I've watched him walk to his phone, turn off the alarm, walk back to bed.
0031  No memory of any of it.
0032  
0033  I went in four times yesterday.
0034  By the third time my voice was different.
0035  By the fourth I was someone I don't want to be.
0036  
0037  His therapist said something I couldn't stop thinking about.
0038  
0039  We keep trying to fix an executive function problem with executive function tools.
0040  
0041  Every alarm, every app, every system I built required him to process, decide, and initiate.
0042  Those are the exact skills ADHD impairs at 6:30am.
0043  
0044  I stopped building better alarm systems.
0045  I started looking for something that doesn't need his executive function at all.
0046  
0047  Dawn Bands vibrates on the wrist.
0048  Touch. Not sound.
0049  A different sensory pathway.
0050  No screen. No dismiss button.
0051  Nothing for his autopilot to turn off.
0052  
0053  He woke up Tuesday before I did.
0054  I was still in bed when I heard him in the kitchen.
0055  
0056  👉 Here's why sound alarms can't work for ADHD brains:
0057  dawnbands.com/pages/adhd-teen-alarm
0058  
0059  60-night guarantee.
0060  Full refund if it doesn't work.
0061  
0062  ---
0063  
0064  ## VARIANT 2 — Objection Killer
0065  **Awareness:** Level 3 (Solution Aware) | **Persona:** Skeptical ADHD Parent/Adult | **Emotion:** Cynicism → Surprise
0066  
0067  **Headline:** I've Spent $300 On Alarms. He Slept Through Every One. Then I Tried This.
0068  
0069  Won't he just sleep through this too?
0070  
0071  That was my first thought.
0072  
0073  My son has ADHD.
0074  I have a drawer full of things that were supposed to fix this.
0075  
0076  He solved the math puzzle app in his sleep.
0077  No memory.
0078  
0079  He walked across the room to his phone, turned it off, walked back.
0080  Eyes open the entire time.
0081  His sleeping body just handles it.
0082  
0083  So another alarm?
0084  Another thing that vibrates?
0085  Another $39 I'll never get back?
0086  
0087  Here's what I didn't understand.
0088  
0089  Every single thing I'd tried uses sound.
0090  Sound goes through one pathway in the brain.
0091  ADHD suppresses that pathway during deep sleep.
0092  
0093  It doesn't matter how loud it is.
0094  It doesn't matter how many you set.
0095  The door is closed.
0096  
0097  This one doesn't use sound.
0098  It vibrates directly on the wrist.
0099  Touch uses a different pathway entirely.
0100  One his brain doesn't shut down during sleep.
0101  
0102  There's no screen.
0103  No app.
0104  Nothing to dismiss.
0105  It just vibrates until he's conscious.
0106  
0107  I ordered it expecting to send it back.
0108  That was eleven weeks ago.
0109  He has not missed a single morning.
0110  
0111  👉 I wrote up what actually changed:
0112  dawnbands.com/pages/adhd-teen-alarm
0113  
0114  60 nights to try it.
0115  Full refund if it doesn't work.
0116  I didn't send it back.
0117  
0118  ---
0119  
0120  ## VARIANT 3 — Pain-First → Mechanism Reveal
0121  **Awareness:** Level 2 (Problem Aware) | **Persona:** ADHD Parent who wants to understand | **Emotion:** Confusion → Clarity
0122  
0123  **Headline:** His Medication Wears Off at 4pm. By 6:30am He's Unreachable. Nobody Explained That Cycle to Me.
0124  
0125  I go into his room at 6:30.
0126  I say his name.
0127  Nothing.
0128  
0129  I go back at 6:35.
0130  Louder.
0131  He grunts.
0132  
0133  6:42.
0134  I'm pulling his covers.
0135  He says he's up.
0136  He is not up.
0137  
0138  6:51.
0139  I'm a different person now.
0140  The voice I swore I wouldn't use.
0141  The words I'll replay in the car.
0142  
0143  He finally moves.
0144  Both of us wrecked before 7am.
0145  
0146  His ADHD medication wears off by 4pm.
0147  By midnight he crashes hard.
0148  By 6:30am he's in the deepest sleep of the entire night.
0149  
0150  That's the window.
0151  That's when every alarm goes off.
0152  And that's when his brain is least capable of hearing any of them.
0153  
0154  Nobody explained this to me.
0155  
0156  Sound alarms fail because ADHD brains close the door on sound during deep sleep.
0157  It's not volume.
0158  It's not willpower.
0159  The signal doesn't reach him.
0160  
0161  Touch is a different door.
0162  One that stays open.
0163  
0164  Dawn Bands vibrates on the wrist.
0165  No sound to filter.
0166  No screen to dismiss in his sleep.
0167  No executive function required.
0168  
0169  Just a physical sensation his brain can't block.
0170  
0171  Three weeks in.
0172  He's up before I walk past his room.
0173  Most mornings I don't even know he's awake until I hear the shower.
0174  
0175  👉 Here's the full explanation of why this works for ADHD:
0176  dawnbands.com/pages/adhd-teen-alarm
0177  
0178  $39.
0179  60-night guarantee.
0180  Works the first night.
0181  
0182  ---
0183  
0184  ## VARIANT 4 — Consequence/Future Fear
0185  **Awareness:** Level 2-3 | **Persona:** ADHD Mom with older teen | **Emotion:** Fear → Urgency
0186  
0187  **Headline:** He Turns 18 in Two Years. Nobody Is Going to Stand Outside His Dorm Room.
0188  
0189  My son is 16 and he has never once woken up on his own.
0190  
0191  Not once.
0192  In sixteen years.
0193  
0194  With ADHD it's not getting better.
0195  The medication cycle makes 6:30am the worst moment of his day.
0196  Meds wore off hours ago.
0197  His brain is in a trough.
0198  Every alarm is hitting a wall.
0199  
0200  I've been the thing that gets him up.
0201  Every morning.
0202  For years.
0203  
0204  And I do it with frustration because by the fourth time I walk in there, I'm not the mom I want to be.
0205  
0206  He turns 18 in two years.
0207  Nobody is going to stand outside his dorm room.
0208  
0209  His roommate is not going to shake him awake.
0210  His professor is not going to wait.
0211  His future boss is not going to call four times.
0212  
0213  I keep thinking about the adults in ADHD forums who say they've lost jobs over this.
0214  Not performance.
0215  Attendance.
0216  Because there was no one left to wake them up.
0217  
0218  That's my son in five years if nothing changes.
0219  
0220  I can't keep being his alarm clock.
0221  But I can't just stop either.
0222  He'll miss everything.
0223  
0224  I needed something external.
0225  Something that works before his medication kicks in.
0226  Something that doesn't require him to process or decide or initiate.
0227  
0228  Dawn Bands vibrates on the wrist.
0229  Touch, not sound.
0230  No screen to dismiss.
0231  No executive function needed.
0232  
0233  It reaches him in the window where nothing else can.
0234  
0235  First week he was up before me every day.
0236  I didn't go in once.
0237  
0238  👉 Here's why this works before medication kicks in:
0239  dawnbands.com/pages/adhd-teen-alarm
0240  
0241  60 nights to try it.
0242  The clock is already running on everything else.
0243  
0244  ---
0245  
0246  ## VARIANT 5 — Risk Reversal + Social Proof
0247  **Awareness:** Level 4 (Product Aware) | **Persona:** ADHD parent/adult ready to try | **Emotion:** Confidence → Action
0248  
0249  **Headline:** 9,000+ Reviews. 60-Night Guarantee. The Only Alarm Built For ADHD Brains.
0250  
0251  ⭐⭐⭐⭐⭐
0252  "ADHD son, 15. Walked to his phone and turned off 3 alarms in his sleep every morning. Zero memory. This band has no screen. Nothing to turn off. He's been up on his own for a month."
0253  — Laura S.
0254  
0255  ⭐⭐⭐⭐⭐
0256  "Diagnosed at 31. I've lost two jobs because of mornings. Set 11 alarms, slept through all of them. This vibrates on my wrist and I'm just awake. Should have had this ten years ago."
0257  — Ryan M.
0258  
0259  ⭐⭐⭐⭐⭐
0260  "His doctor asked what changed with the mornings. I said a vibrating wristband. She just nodded and said 'keep using it.' First time a solution actually stuck."
0261  — Danielle T.
0262  
0263  9,000+ reviews.
0264  $39.
0265  60-night money-back guarantee.
0266  
0267  If you have ADHD or love someone who does, you already know.
0268  
0269  Sound alarms don't work.
0270  Puzzle apps don't work.
0271  Louder doesn't work.
0272  More doesn't work.
0273  
0274  The reason is simple.
0275  ADHD brains suppress sound during deep sleep.
0276  Every alarm you've tried hits the same closed door.
0277  
0278  Dawn Bands vibrates on the wrist.
0279  Touch.
0280  A different pathway.
0281  No screen.
0282  No dismiss button.
0283  Nothing that needs executive function to process.
0284  
0285  Try it for 60 nights.
0286  Full refund if it doesn't work.
0287  No questions. No hoops.
0288  
0289  Most people see it work on the first night.
0290  
0291  👉 Here's why it works for ADHD when nothing else does:
0292  dawnbands.com/pages/adhd-teen-alarm

Native copy / assets/gold-standard-example.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/assets/gold-standard-example.md

SHA-256: ea237599482d40993e7a0d336c5910c0bbb832181b68fcd1cc12c68f5fe2a056

Full source · 507 lines
0001  # Native8 — Teen Mom Listicle — Test Batch (FINAL)
0002  **Based on:** native8 — $15.25 CPP / 3.04x ROAS / 17.6% CTR (proven winner)
0003  **Destination URL:** dawnbands.com/pages/7-reasons-teen-sleeps-through-alarms
0004  **Date:** 2026-03-07
0005  
0006  > **Use for structure and cadence, not current product truth or proof.** Before adapting any line, replace price, offer, setup, destination, mechanism, and outcome claims from the current validated brief and canonical product-truth file. Historical first-person outcomes in this sample are not reusable proof unless independently sourced for the new ad.
0007  
0008  ---
0009  
0010  ## VARIANT 1 — The Trash Bag (LONG, 500-700 words)
0011  **Awareness:** Level 3 | **Persona:** Mom who snapped | **Emotion:** Dark satisfaction → explanation → relief
0012  **Mechanism Delivery:** Analogy (locked door)
0013  
0014  My husband thought I lost my mind.
0015  
0016  I was standing in our son's room at 6:47am holding a trash bag.
0017  
0018  Not the white kitchen kind.
0019  The big black contractor bag.
0020  
0021  I'd already knocked four times.
0022  Called his name six.
0023  Went back downstairs twice.
0024  Came back up screaming once.
0025  
0026  He was asleep through all of it.
0027  
0028  His phone had seven alarms set.
0029  I could hear every single one from the hallway.
0030  He couldn't hear them from two feet away.
0031  
0032  So I started putting his stuff in the bag.
0033  Xbox controller.
0034  Headphones.
0035  The shoes he'd begged me for.
0036  
0037  My husband walked in and looked at me like I needed help.
0038  
0039  "What are you doing?"
0040  
0041  I didn't answer him.
0042  I didn't have words for what I was doing.
0043  I was a 43-year-old woman filling a trash bag at 6am because I'd become someone I didn't recognize.
0044  
0045  The morning turned me into a yelling banshee.
0046  That's what I called myself.
0047  Because that's what I was.
0048  
0049  Every single morning.
0050  For three years.
0051  
0052  I tried the Sonic Boom alarm.
0053  He slept through it.
0054  I tried putting his phone across the room.
0055  He walked to it in his sleep.
0056  Eyes open. Turned it off. Walked back to bed.
0057  No memory of doing it.
0058  
0059  I tried the math puzzle app.
0060  He solved them unconscious.
0061  His sleeping self defeats everything.
0062  
0063  I tried cold water.
0064  Taking his blankets.
0065  Opening every blind.
0066  Calm conversations about responsibility.
0067  Consequences.
0068  Rewards.
0069  
0070  None of this works.
0071  I've tried all of this and more.
0072  
0073  After the trash bag morning, I sat in my car in the school parking lot and googled "why can't my teenager wake up to alarms."
0074  
0075  Not for the first time.
0076  But this time I actually read something that made sense.
0077  
0078  It said sound alarms don't fail because they're too quiet.
0079  They fail because the brain locks the door on sound during deep sleep.
0080  
0081  Locked.
0082  Not cracked.
0083  Not slightly closed.
0084  Locked.
0085  
0086  You can bang on a locked door all morning.
0087  You can scream through it.
0088  You can set seven alarms to bang on it for you.
0089  
0090  The door does not open.
0091  
0092  It doesn't matter how loud.
0093  It doesn't matter how many.
0094  The auditory pathway shuts down in deep sleep, and every alarm on earth uses that same pathway.
0095  
0096  That's why he walks to his phone asleep and turns it off.
0097  His body hears enough to react.
0098  His brain never wakes up.
0099  
0100  The thing I found uses touch instead of sound.
0101  Vibrates on the wrist.
0102  Touch goes through a completely different door.
0103  One that doesn't lock during sleep.
0104  
0105  No screen. No app. No buttons.
0106  It just vibrates until he's actually conscious.
0107  
0108  I didn't tell my husband I ordered it.
0109  I didn't tell my son.
0110  I put it on his wrist Sunday night and set it for 6:15.
0111  
0112  Monday morning I woke up to my own alarm.
0113  Not his.
0114  Not yelling.
0115  My alarm.
0116  
0117  I lay there and waited.
0118  I heard footsteps.
0119  His footsteps.
0120  Before I even got out of bed.
0121  
0122  I waited for it to fail on Tuesday.
0123  It didn't.
0124  
0125  Wednesday I was sure it was over.
0126  It wasn't.
0127  
0128  It's been nine weeks.
0129  I have not gone into his room once.
0130  
0131  The trash bag is still in my closet.
0132  His stuff is back in his room.
0133  He doesn't know why I took it.
0134  I don't want him to.
0135  
0136  I don't yell anymore.
0137  The first version of me he sees isn't angry.
0138  That's it.
0139  That's all I wanted.
0140  
0141  👉 Here's what I read:
0142  dawnbands.com/pages/7-reasons-teen-sleeps-through-alarms
0143  
0144  I had 60 nights to send it back.
0145  Never even considered it.
0146  
0147  **Headline:** I filled a trash bag with his stuff at 6am. That was the morning I stopped using sound alarms.
0148  
0149  ---
0150  
0151  ## VARIANT 2 — The Apology / 847 Mornings (LONG, 500-700 words)
0152  **Awareness:** Level 3 | **Persona:** Mom reckoning with guilt | **Emotion:** Guilt → understanding → forgiveness
0153  **Mechanism Delivery:** Full science
0154  
0155  847 mornings.
0156  
0157  I counted.
0158  
0159  From the September he started high school to the Tuesday I finally understood what was happening, that's how many mornings I went into his room and turned into someone I don't want to be.
0160  
0161  847 mornings of knocking.
0162  Then calling.
0163  Then louder.
0164  Then yelling.
0165  Then both of us wrecked before 7am.
0166  
0167  I was a human alarm clock.
0168  And the worst part is the awake version of my son does not remember the previous 9 attempts.
0169  Only the rageful 10th one.
0170  
0171  That's the only version of me he knows in the morning.
0172  Angry. Loud. Done.
0173  
0174  I wake up before my alarm now.
0175  Not because I'm rested.
0176  Because my body knows what's coming.
0177  I lie there calculating how much buffer I have before I have to start.
0178  
0179  Among the many hats I wear, the one that fits tightest is NAG.
0180  I don't want him to wake up to my anger.
0181  But I don't know how to wake him up without it.
0182  
0183  His phone has seven alarms.
0184  They've woken ME up before they wake him.
0185  He shuts them off and goes back to sleep.
0186  
0187  I put the phone across the room.
0188  He walked to it in his sleep.
0189  Eyes open the whole time. Turned it off. Walked back.
0190  No memory.
0191  
0192  I demand to hear evidence he is sitting up and out of bed.
0193  He says he is.
0194  Repeat 10 times over half an hour.
0195  His sleeping self tells me he is awake.
0196  
0197  I tried everything.
0198  Sonic Boom alarm.
0199  Three Alexas.
0200  Math puzzle apps he solved unconscious.
0201  Cold water.
0202  Lights on.
0203  Taking his phone.
0204  Consequences.
0205  Incentives.
0206  Calm conversations.
0207  
0208  None of this works.
0209  I've tried all of this and more.
0210  
0211  Last February his English teacher pulled me aside at conferences.
0212  
0213  "Is everything okay at home? He seems checked out in first period. Every day."
0214  
0215  I almost cried in the hallway.
0216  
0217  That was the night I read something that finally explained it.
0218  
0219  The teenage brain suppresses the auditory processing pathway during deep sleep.
0220  It's not laziness.
0221  It's not defiance.
0222  It's neurology.
0223  
0224  Sound alarms use one sensory channel to reach the brain.
0225  During deep sleep, that channel shuts down.
0226  Completely.
0227  
0228  It doesn't matter how loud the alarm is.
0229  It doesn't matter how many you set.
0230  It doesn't matter if he has to solve equations to dismiss it.
0231  His brain processes enough to react physically.
0232  But never enough to reach consciousness.
0233  
0234  That's why he walks to his phone asleep.
0235  That's why he answers me while unconscious.
0236  That's why nothing has worked.
0237  
0238  Every single thing I tried for 847 mornings used sound.
0239  Every single one went through the same blocked pathway.
0240  I wasn't failing.
0241  The method was.
0242  
0243  Touch uses a different sensory pathway.
0244  One the brain does not suppress during sleep.
0245  Vibration on the skin reaches the brain through a channel that stays open.
0246  
0247  I found a wrist band that vibrates.
0248  No sound. No screen. No app.
0249  Nothing for his sleeping self to dismiss.
0250  It just vibrates on his skin until he's conscious.
0251  
0252  I put it on him on a Sunday.
0253  
0254  Day 1: He was in the kitchen before me.
0255  Day 4: He made himself breakfast.
0256  Day 11: His teacher emailed me.
0257  "Whatever you changed, it's working. He's a different kid in first period."
0258  
0259  She noticed before I even said anything.
0260  
0261  It's been fourteen weeks.
0262  He hasn't missed a single morning.
0263  
0264  But that's not the part that made me cry.
0265  
0266  Last Thursday he came downstairs and said something to me before I said anything to him.
0267  
0268  "Morning, Mom."
0269  
0270  He hadn't done that in over a year.
0271  
0272  I'll put the link below.
0273  
0274  dawnbands.com/pages/7-reasons-teen-sleeps-through-alarms
0275  
0276  $49.
0277  Full refund if it doesn't work within 60 nights.
0278  
0279  847 mornings of doing it wrong.
0280  It took one night of doing it differently.
0281  
0282  **Headline:** 847 mornings I yelled him awake. I wish I'd known it wasn't his fault.
0283  
0284  ---
0285  
0286  ## VARIANT 3 — The Neighbor (LONG, 500-700 words)
0287  **Awareness:** Level 2 | **Persona:** Mom who saw herself from outside | **Emotion:** Humiliation → determination → calm
0288  **Mechanism Delivery:** Someone else explained it (pediatrician)
0289  
0290  "Do you need help? I hear screaming from your house every morning at 7am."
0291  
0292  My neighbor said that to me over the fence on a Saturday.
0293  
0294  A Saturday.
0295  
0296  The one morning I wasn't screaming.
0297  
0298  I laughed it off.
0299  Made some joke about teenagers and alarm clocks.
0300  Walked inside and locked the bathroom door and sat on the floor.
0301  
0302  Because she was right.
0303  
0304  Every morning at 6:45 it starts.
0305  I knock.
0306  I call his name.
0307  I wait.
0308  I go back.
0309  I call louder.
0310  I yell.
0311  
0312  I call him for 2 hours straight some mornings.
0313  
0314  His alarm has woken ME up before it wakes him.
0315  Seven alarms on his phone.
0316  I hear all of them.
0317  He hears none.
0318  
0319  He shuts them off and goes back to sleep.
0320  
0321  I put the phone across the room.
0322  He got up, walked to it, turned it off, walked back to bed.
0323  Eyes open.
0324  Full conversation with me.
0325  No memory of any of it.
0326  
0327  His sleeping self tells me he is awake.
0328  His sleeping self lies.
0329  
0330  I became ridiculously rage-y.
0331  That's the word I used when I told his pediatrician.
0332  I said I'm becoming a person I hate at 6am and I can't stop it.
0333  
0334  She didn't tell me to try harder.
0335  She didn't suggest another alarm app.
0336  
0337  She said something no one had ever explained to me.
0338  
0339  "His brain blocks sound during deep sleep. It's a sensory pathway issue. The auditory channel closes. It doesn't matter how many alarms you set or how loud they are. They all go through the same channel. And for him, that channel is shut."
0340  
0341  I sat there staring at her.
0342  
0343  Three years.
0344  Sonic Boom alarms.
0345  Multiple Alexas in his room.
0346  Math apps he solved in his sleep.
0347  Cold water.
0348  Blinds open.
0349  Lights blazing.
0350  
0351  All sound.
0352  All the same blocked channel.
0353  Every single one.
0354  
0355  She told me about something that uses touch instead.
0356  A band that vibrates on the wrist.
0357  Touch goes through a different pathway.
0358  One that stays open during deep sleep.
0359  
0360  No screen. No buttons. No app.
0361  Nothing for his unconscious body to dismiss.
0362  
0363  I ordered it from the car in the parking lot.
0364  
0365  The first morning I didn't go into his room.
0366  I heard him get up.
0367  I heard the shower.
0368  I stood in the kitchen holding my coffee with both hands and I didn't move.
0369  
0370  The second morning was the same.
0371  The third.
0372  The seventh.
0373  
0374  No screaming.
0375  No knocking.
0376  No third trip upstairs.
0377  No yelling through the door.
0378  No standing over his bed hating who I was becoming.
0379  
0380  Week four, I was outside watering the plants.
0381  My neighbor came to the fence again.
0382  
0383  "It's been quiet over there in the mornings."
0384  
0385  I said yeah.
0386  That's all I said.
0387  
0388  It's been twelve weeks.
0389  He has not missed one morning.
0390  
0391  The explanation is at dawnbands.com/pages/7-reasons-teen-sleeps-through-alarms
0392  
0393  My neighbor doesn't hear me anymore.
0394  Nobody does.
0395  
0396  **Headline:** My neighbor heard me screaming every morning at 7am. She doesn't hear anything now.
0397  
0398  ---
0399  
0400  ## VARIANT 4 — The Drawer (SHORT, 150-300 words)
0401  **Awareness:** Level 3 | **Persona:** Skeptic burned by failed products | **Emotion:** Dark humor → surprise
0402  **Mechanism Delivery:** Barely explained
0403  
0404  I have a junk drawer that costs more than my groceries.
0405  
0406  Sonic Boom alarm. $45. He slept through it.
0407  Phone alarm dock. $30. Slept through it.
0408  Math puzzle app. $12. Solved it in his sleep. No memory.
0409  Sunrise lamp. $55. Decorative.
0410  Three Alexas. $90 total. They wake me up. Not him.
0411  
0412  $232 of things that were supposed to fix mornings.
0413  
0414  All of them use sound.
0415  Turns out his brain shuts sound off during deep sleep.
0416  Didn't matter how loud.
0417  Didn't matter how many.
0418  
0419  He walked across his room to his phone in his sleep.
0420  Eyes open. Turned it off. Walked back.
0421  Zero memory.
0422  
0423  His sleeping body defeats everything.
0424  
0425  This one vibrates on his wrist.
0426  No sound. No screen. Nothing to dismiss unconscious.
0427  
0428  I ordered it expecting it to go in the drawer.
0429  
0430  That was ten weeks ago.
0431  
0432  It's not in the drawer.
0433  
0434  He gets up before I go in.
0435  Every morning.
0436  I haven't yelled since November.
0437  
0438  Link is here if you want to read it:
0439  dawnbands.com/pages/7-reasons-teen-sleeps-through-alarms
0440  
0441  $49. 60 nights. Full refund if it joins the drawer.
0442  
0443  It didn't.
0444  
0445  **Headline:** $232 in that drawer. $49 on his wrist. Guess which one works.
0446  
0447  ---
0448  
0449  ## VARIANT 5 — The Car (SHORT, 150-300 words)
0450  **Awareness:** Level 4 | **Persona:** Mom in school parking lot | **Emotion:** Raw → relief
0451  **Mechanism Delivery:** Skip entirely (outcome only)
0452  
0453  He got out of the car without looking at me.
0454  
0455  Didn't say bye. Didn't say anything.
0456  
0457  I sat in the school parking lot for ten minutes replaying everything I'd said that morning.
0458  
0459  The knocking.
0460  The yelling.
0461  The "I've told you TEN TIMES."
0462  The door slam.
0463  
0464  The first version of me he sees every day is angry.
0465  That's the only mom he knows before 8am.
0466  
0467  I was a human alarm clock.
0468  And the alarm was my voice at full volume.
0469  
0470  I don't want him to wake up to my anger.
0471  But I didn't know how else to get him up.
0472  
0473  Seven phone alarms.
0474  He sleeps through them.
0475  I hear every one.
0476  
0477  I tried everything for two years.
0478  Nothing worked.
0479  I gave up trying to fix it and started trying to survive it.
0480  
0481  Someone posted this in a mom group I'm in.
0482  I almost scrolled past it.
0483  
0484  I didn't expect anything.
0485  
0486  First morning: he was up before I went in his room.
0487  Second morning: same.
0488  Third morning: I heard him in the shower.
0489  
0490  It's been eight weeks.
0491  
0492  Yesterday he got out of the car and turned around.
0493  
0494  "Bye, Mom."
0495  
0496  Two words.
0497  I cried in the parking lot again.
0498  Different reason.
0499  
0500  Someone posted this in my group:
0501  dawnbands.com/pages/7-reasons-teen-sleeps-through-alarms
0502  
0503  P.S. He doesn't know I used to sit in that parking lot after drop-off.
0504  He doesn't know I cried.
0505  He just knows mornings are different now.
0506  
0507  **Headline:** He stopped saying bye to me. That was the part that broke me.

Native copy / references/adversarial-batch-copy-audit.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/adversarial-batch-copy-audit.md

SHA-256: 7432cee3f45b7baff3938152d73583ba76279c1849767e8b8e41ff0af0379153

Full source · 99 lines
0001  # Adversarial Batch Copy Audit Protocol
0002  
0003  Use this when auditing a finished native-ad batch against structured briefs and VOC receipts. This is an editorial/provenance audit, not a deterministic artifact check.
0004  
0005  ## Core distinction
0006  
0007  A deterministic PASS proves file count, required fields, word ranges, formatting, offer order, destinations, and banned-term rules. It does **not** prove that a first-person story is true, that precise details are sourced, that the avatar voice is faithful, or that twenty ads are genuinely diverse.
0008  
0009  Always report these as separate results:
0010  
0011  1. **Artifact-contract status** — objective checks.
0012  2. **Editorial/provenance status** — hook, voice, persuasion, truth, and diversity.
0013  
0014  Never let artifact PASS launder an editorial FAIL.
0015  
0016  ## Evidence hierarchy
0017  
0018  For every first-person scene, quotation, outcome, time, count, price, duration, product trial, or external witness:
0019  
0020  1. Find an exact source receipt for the same speaker and fact.
0021  2. If only the behavior or emotion is sourced, the assembled narrator is a **composite**, not a testimonial.
0022  3. A creative brief authorizing an artifact (for example, “five-day log” or “DORM box”) does not prove that the event happened.
0023  4. Plausibility is not provenance.
0024  5. Safe repairs are:
0025     - attribute exact VOC (`A parent wrote…`);
0026     - frame it explicitly as a hypothetical (`Picture…`, `Before you pack…`, `Try writing…`);
0027     - use a non-testimonial ownership argument;
0028     - remove unsupported precision.
0029  
0030  ### Precision test
0031  
0032  Treat specificity as proof-shaped. Numbers such as “seven minutes,” “three problems,” “six steps,” or a minute-by-minute sequence require direct receipts. If the exact number is unsupported, remove it rather than replacing it with a different invented number.
0033  
0034  ## Required scoring layers
0035  
0036  Score every ad independently from 1–10 on:
0037  
0038  - Hook WTF factor
0039  - Hook specificity
0040  - Hook curiosity gap
0041  - Lines 3–8 specificity
0042  - AI-sounding language, abstraction, and filler
0043  - Awareness/sophistication alignment
0044  - Resignation moment, villain, root cause, and mechanism chain
0045  - Avatar fidelity against exact VOC
0046  - Victory as absence
0047  - Claim and testimonial integrity
0048  - CTA and body/headline/description division
0049  - Cross-ad distinctiveness
0050  
0051  Anything below 7 must include exact proposed replacement line(s), not only diagnosis. If the task is a read-only audit, label replacements as proposals and do not edit source ads.
0052  
0053  ## Blocking vs taste
0054  
0055  ### High-confidence blockers
0056  
0057  - Invented or composite first-person experience presented as literal testimony
0058  - Unsupported outcome, chronology, dialogue, time, count, or external witness
0059  - Medical/every-time claims beyond approved proof
0060  - Wrong awareness level or missing governing objection
0061  - Repeated body/mechanism blocks that collapse supposedly distinct ads
0062  - Failure to satisfy an explicit brief requirement
0063  
0064  ### Subjective taste flags
0065  
0066  - Hook preference when multiple options are defensible
0067  - Degree of quietness, humor, or emotional intensity
0068  - Stylistic wording where truth and brief alignment already pass
0069  - Optional field-role improvements
0070  
0071  Always label confidence so editorial preference is not presented as proof failure.
0072  
0073  ## Cross-ad diversity check
0074  
0075  Run both exact and conceptual checks:
0076  
0077  - repeated full sentences;
0078  - repeated hook formulas;
0079  - repeated mechanism cadence;
0080  - repeated victory artifacts (door, footsteps, bathroom water, hot coffee);
0081  - narrator sameness across parent, teen, partner, or other personas;
0082  - repeated description/headline restatement.
0083  
0084  Assign each ad one proof artifact or mechanism image it uniquely owns. Repetition required by canon (for example, an exact guarantee token order) is less concerning than repeated narrative prose.
0085  
0086  ## Performance boundary
0087  
0088  Historical performance can support hypotheses about territories or devices. It cannot prove that a current hook, angle, or wording is a winner. State stale/current boundaries explicitly.
0089  
0090  ## Deliverable verification
0091  
0092  Before delivery verify:
0093  
0094  - every expected ad has one score row and one detailed section;
0095  - every sub-7 score has exact proposed text;
0096  - every blocker includes snippet, rationale, and confidence;
0097  - invented testimonial risk is explicitly addressed;
0098  - high-confidence blockers are separated from taste flags;
0099  - no source ad was edited when the assignment was read-only.

Native copy / references/alevia-live-anatomy-2026-08-12.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/alevia-live-anatomy-2026-08-12.md

SHA-256: e82e0c80d37209d837f908a3e552eed8657b2547e912364719d59aeed619ec7b

Full source · 96 lines
0001  # Alevia live-account anatomy — measured update (2026-08-12)
0002  
0003  Source: 240 unique active ad copies sampled across all 14 Alevia pages (TrendTrack domain query,
0004  last-3m actives), raw JSON at `~/.openclaw/workspace/output/trendtrack-alevia-2026-08-12/`.
0005  Full analysis: `~/.openclaw/workspace/knowledge/brands/dawnbands/alevia-native-anatomy-2026-08-12.md`.
0006  
0007  This file UPDATES the two-control evidence base that `dawnbands-long-form-witness-native.md` was
0008  built from (those controls: ~3,188/3,662 words, ~60% reveal). It is measurement, not new contract:
0009  where a number below differs from the Mode B contract, the contract stays default and the measured
0010  band is a SANCTIONED TEST VARIABLE. Provenance, narrator-authority, and claim-ceiling gates are
0011  untouched by everything here.
0012  
0013  ## Measured distributions (n=240 unless noted)
0014  
0015  - **Length:** p10 2,373 / median 3,411 / p90 4,338 words. 229/240 are 1,500+. Our 2,500–4,000
0016    contract sits inside their live band; their upper half (4,000–5,000) is unexplored by us.
0017  - **Reveal position:** median 52%, p10 31%, p90 66%. The two July controls (~60%) sit high in
0018    the live distribution. Contract stays 55–70%; **45–55% is a sanctioned test band**, not a defect.
0019  - **Formatting (settles the line-format question):** paragraph prose, median ~31 words/paragraph
0020    (2–4 sentences). 133/240 mix SHORT staccato beats at punch moments only. Zero copies are
0021    wall-to-wall one-sentence-per-line. This matches the 2026-07-31 Mode B ruling (PARAGRAPH prose)
0022    and supersedes the older sentence-per-line formatting block for Mode B. Sentence-per-line
0023    remains correct for Mode A / compressed natives and static primaries.
0024  - **Register hygiene (long-form):** emoji 12/240 (excluding the final 👉 link arrow), ALLCAPS
0025    lines 6/240, comment-bait 25/240. The witness lane is kept clean of ad-register signals.
0026  - **Headline field:** 0/240 witness natives use the Meta title field. Observation only — the
0027    5-headlines-per-ad rule stands; a blank/minimal-title witness cell is a legitimate test.
0028  
0029  ## Opener — measured ceiling of the rupture-object cold open
0030  
0031  All 20 longest-running copies (95–127d) open the same way: **mundane object + weekday + month +
0032  place-name in sentence one, confession register, zero product/category scent.**
0033  
0034  > "I pulled my mother of the bride dress out of the garment bag on a Tuesday afternoon in March…"
0035  > "The Christmas card I mailed to my cousin Patty in November came back… a yellow sticker that said DECEASED."
0036  > "My dad couldn't open the register at the pizza shop last Tuesday… the same brass-plated NCR since 1987."
0037  
0038  Audit line for every Mode B opening: **could this first sentence appear in a personal Facebook
0039  post with zero ad intent?** If it reads like a hook, it fails.
0040  
0041  Provenance caveat: that level of specificity (weekday, month, place, object) requires `RECEIPT`
0042  or explicit `AUTHORIZED DRAMATIZATION` per the provenance ledger. The formula does not license
0043  invented precision — it tells you what the receipt-backed or authorized detail should LOOK like
0044  when you have it.
0045  
0046  ## Reveal — the spoken-word mechanic
0047  
0048  The product never enters as a decision. In every measured survivor it enters as a **word spoken
0049  by a named third party inside the story**, often at the end of a narrated discovery chain:
0050  
0051  > "She told us about a fruit called Amla." · "Renée told me about Amla." ·
0052  > "…a thread on a health forum… a comment from a woman named Patrice…" ·
0053  > "I found my way to amla through a Reddit thread… written by a woman in Ohio…"
0054  
0055  Two refinements for Beat 6/9 of the witness architecture:
0056  1. **Discovery chain**: article → forum thread → named stranger's comment → skeptical read → try.
0057     Narrate the chain itself; it does DR work (relatability + implied social proof without a claim).
0058  2. **Category-then-brand split**: mid-story reveal names the CATEGORY/mechanism object (for us:
0059     the wrist-worn silent alarm as a thing someone mentions), the BRAND surfaces only at the
0060     link/close. Never brand-name the mid-story reveal.
0061  
0062  Both routes obey the existing gates: the named stranger is dialogue inside the story (allowed),
0063  never an authority byline or a fabricated named testimonial with product outcomes (banned).
0064  
0065  ## Close — the P.S. ladder (measured default, our biggest structural gap)
0066  
0067  217/240 copies close with stacked postscripts, up to P.P.P.P.S. Each rung has a distinct job:
0068  
0069  1. **P.S. — rupture-object return.** ("The dress is hanging in the closet now…")
0070  2. **P.P.S. — guarantee naturalized in-voice.** ("The 90-day money-back guarantee is real. I asked.")
0071     For DawnBands: only the verified refund terms, usually 60-night money-back; all offer gates apply.
0072  3. **P.P.P.S. — objection/placebo kill in-voice.** ("Someone is going to say 'it's just placebo.'…")
0073  4. **Final P.S. — reader-direct-address AS the CTA.** ("If you're the husband reading this over
0074     your wife's shoulder… don't make her wait for the ambulance.") For us: the mom reading at 1 AM
0075     with tomorrow's alarm already set.
0076  5. **Bare 👉 link on its own line AFTER the final coda.** No CTA copy around it.
0077  
0078  Difference vs the existing close architecture (motif → loss → role → offer → CTA, codas optional):
0079  in the measured winners the codas ARE the close and the link comes last. Treat the ladder as
0080  **close variant B** — test it against the existing close on a proven copy before making it default.
0081  The existing outcome-timeline, offer-verification, and desired-absence gates apply to every rung.
0082  
0083  ## Device frequencies (angle-selection input, not copy instructions)
0084  
0085  Of 95 primary-page copies: spouse-witness 57 · death-stakes 54 · family-milestone 41 ·
0086  wrong-diagnosis reveal 25 · authority-defiance 25 · insider-authority 14. Failed-solution stack
0087  with prices 70/95. DawnBands transfer analysis + which of these are banned for us (death-ledes,
0088  fabricated medical authority) lives in
0089  `~/.openclaw/workspace/knowledge/brands/dawnbands/alevia-application-recs-2026-08-12.md`.
0090  
0091  ## Format-page discipline (context for deployment, not copy)
0092  
0093  Alevia never crosses formats across page identities: witness natives on community/persona pages,
0094  55–84-word ✅-checklist shorts on brand pages (their longest-running ads, 118–127d — see T4 in
0095  `avelia-short-native-templates.md`), video quarantined on one page with ~3 copy skeletons.
0096  Our page axis in the ad-batch-hub matrix now tracks the same variable.

Native copy / references/avelia-short-native-templates.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/avelia-short-native-templates.md

SHA-256: 4f593a181226ce6242f99aefeebf6a2b09d2a114c3ca8d9fe90106d418bfd767

Full source · 140 lines
0001  # Avelia-derived short native template system
0002  
0003  Use this reference inside the class-level `native-copy` workflow when analyzing competitor short-form native ads or drafting compressed Meta copy from trigger-specific, testimonial, or failed-solution entry points.
0004  
0005  ## Evidence boundary
0006  
0007  - **Observed:** exact advertiser/page/shop/domain fields, ad IDs, body/card fields, dates, statuses, counts, and null metrics from source receipts.
0008  - **Inferred:** awareness mapping, persuasion rationale, recommended word ranges, template selection, and adaptation advice.
0009  - Duplicate count, longevity, and zero/null reach are not performance proof.
0010  - Source claims never transfer to the target brand without independent verification.
0011  
0012  ## Required inputs
0013  
0014  Brand/product/destination; exact avatar context; awareness; root cause → product mechanism → differentiator; verified claims and sources; quantified facts or explicit empty list; guarantee/effort/timing; claim restrictions; requested fields.
0015  
0016  Stop if mechanism, context, proof inventory, restrictions, or destination-awareness fit are missing. Never invent ingredients, statistics, timelines, guarantees, dosage, clinical status, or outcomes.
0017  
0018  ## Workflow
0019  
0020  1. In analysis mode, verify advertiser identity, page ID, shop/domain, destination, exact ad IDs, and direct receipts. Preserve nulls and source field semantics.
0021  2. Name what the reader already tried or believes; select one narrow trigger/context.
0022  3. Choose exactly one template below. Do not hybridize until each passes independently.
0023  4. Label skeleton beats `[FACT]`, `[VOC]`, `[INFERENCE]`, or `[PROHIBITED]`; delete prohibited beats.
0024  5. Assign field roles before prose:
0025     - primary text = tension + mechanism + product bridge + close;
0026     - headline = concise angle only when the target platform field exists;
0027     - description = proof, guarantee, effort, or quantified differentiation;
0028     - CTA = awareness- and register-matched.
0029  6. Run claim, specificity, field-complementarity, and cross-variant diversity QA.
0030  7. Return every requested field plus a verification receipt.
0031  
0032  ## T1 — Trigger-Specific Diagnosis → Mechanism → Compressed Remedy
0033  
0034  Best for problem-aware readers who recognize a life trigger but not the mechanism.
0035  
0036  ```text
0037  [Specific life event or sustained context].
0038  Now [visible symptom/behavior], and [betrayal or nobody-warned-me gap].
0039  
0040  [Verified prevalence, threshold, or mechanism].
0041  It's not [default explanation].
0042  [Cost of delay or why prior assumption fails].
0043  
0044  [Product] was designed/formulated for [verified mechanism].
0045  [Low-effort regimen].
0046  [Verified result framing or non-outcome benefit].
0047  
0048  [Resolve the opening tradeoff].
0049  ```
0050  
0051  Observed Avelia range: 85–96 words; product appeared 57–62% into body words. Treat this as evidence, not a mandatory threshold. Inferred reusable target: 80–110 words.
0052  
0053  ## T2 — Result Confession → Authority Bridge → Benefit Stack
0054  
0055  Best for solution-/product-aware readers needing a compact result and reasons to believe.
0056  
0057  ```text
0058  This [format]? [Natural verdict, no cliché] for [specific result].
0059  I've noticed [detail 1], [detail 2], and [detail 3], even during [adverse context].
0060  
0061  [Brand/product] uses [verified mechanism] to [verified action].
0062  [Connect format to convenience or desired state].
0063  
0064  • [Benefit 1]
0065  • [Benefit 2]
0066  • [Benefit 3 / objection handling]
0067  
0068  [Soft, specific desired-state CTA].
0069  ```
0070  
0071  Observed Avelia length: 106 words. Inferred reusable target: 90–120. Never leave placeholders or use “game-changer.”
0072  
0073  ## T3 — Failed Category Reframe → Quantified Mechanism Stack
0074  
0075  Best for skeptical solution-aware readers who tried the category and felt nothing.
0076  
0077  ```text
0078  You didn't need MORE [category].
0079  You needed the right [verified mechanism variable].
0080  
0081  [Product]: [count/specification], [dose/time], [convenience], designed to [verified benefit domain].
0082  ```
0083  
0084  Observed Avelia range: 32–46 words. Inferred reusable target: 30–55.
0085  
0086  ## T4 — Branded Checklist Evergreen (added 2026-08-12, measured from live actives)
0087  
0088  Best for product-/offer-aware traffic on the BRAND page only (never the community/persona
0089  pages). Measured basis: Alevia's longest-running ads of any format (118–127 days live) are
0090  54–84-word checklist shorts on their brand-identity pages; their new product lanes launch on
0091  the same skeleton unchanged (`alevia-live-anatomy-2026-08-12.md`).
0092  
0093  ```text
0094  [Product/category], the [one-line identity + verified mechanism clause], naturally/simply.
0095  
0096  ✅ [Benefit 1 — verified or non-outcome]
0097  ✅ [Benefit 2]
0098  ✅ [Benefit 3]
0099  ✅ [Benefit 4 / objection kill]
0100  
0101  [One heritage/credibility line IF verifiable — otherwise omit].
0102  
0103  Try [product] risk-free for [verified refund window] or get 100% of your money back[, no
0104  questions asked — only if the verified policy says so].
0105  ```
0106  
0107  Observed range: 54–84 words. Emoji checklist is the register for THIS template (it is banned
0108  in long-form witness copy). Every benefit line passes the claim gates; the refund line uses
0109  only currently verified policy terms (DawnBands: 60-night money-back, $49 only if current).
0110  Role: the evergreen BOF/retarget floor — build 3–4, leave them running, never on story pages.
0111  
0112  ## Anti-patterns
0113  
0114  - Same-name advertiser mixing without identity verification.
0115  - Relabeling card descriptions as platform headlines.
0116  - Unsupported clinical claims, prevalence, dosage, timelines, “no side effects,” or guaranteed outcomes.
0117  - Multiple root causes in one short ad.
0118  - Generic symptom hook when a trigger-specific hook exists.
0119  - Repeating the same claim across primary text, headline, and description.
0120  - Importing early reveal timing into a long-form cold story.
0121  - Placeholders, clichés, decorative emoji clutter, dense paragraphs, or direct SHOP NOW language that breaks a quiet register.
0122  
0123  ## Verification receipt
0124  
0125  ```markdown
0126  - Template: T1 | T2 | T3
0127  - Awareness: ...
0128  - Product reveal: sentence N / approximately X%
0129  - Primary text words: N
0130  - Field completeness: primary/headline/description/CTA/URL
0131  - Claims checked against: [sources]
0132  - Unsupported claims removed: [list]
0133  - Hook/body/card relationship: [one sentence]
0134  - QA result: PASS | FAIL
0135  - Remaining uncertainty: [list]
0136  ```
0137  
0138  ## Evidence note
0139  
0140  Session evidence dated 2026-07-18 contained 62 deduplicated Avelia/Avelia Care ads and 9 exact primary-text variants. Two current-page ads were directly scanned active; the remainder were linked-page records. Numeric reach/scaling evidence was absent. This reference preserves structures, not winner claims.

Native copy / references/consolidated-native-ad-repair-pass.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/consolidated-native-ad-repair-pass.md

SHA-256: af41b77f2bec29009f99b17288673f6ae59bb77a23af1783465d8549d22fb80f

Full source · 73 lines
0001  # Consolidated Native-Ad Repair Pass
0002  
0003  Use this when completed ads have multiple read-only QA reports and the user requests one repair plus revalidation pass.
0004  
0005  ## Authority and scope
0006  
0007  1. Re-read the live ad files immediately before editing.
0008  2. Extract only the requested brief objects and their exact VOC receipts.
0009  3. Consolidate findings from formal QA, sharpening/avatar review, deterministic audit, and any later SOP/performance overlay.
0010  4. Carry forward every unresolved blocker unless a later report explicitly retires it.
0011  5. Edit only the authorized ad files and manifest. Do not create a new audit report unless requested.
0012  
0013  ## Repair order
0014  
0015  1. **Truth first:** remove composite narrators, invented chronology, precise unsupported times/counts, and fabricated dialogue.
0016  2. **Quote fidelity:** text in quotation marks must match one receipt exactly. Otherwise remove quotation marks and attribute the paraphrase. An excerpt must remain a literal source substring: never add a period inside the closing quote when the source sentence continues. End the exact fragment without invented punctuation, include the full source sentence, or paraphrase it unquoted.
0017  3. **Brief fidelity:** restore assigned mechanism delivery, structure, awareness, objection, word range, CTA, and destination.
0018  4. **Craft:** sharpen hooks, remove strategy-deck language, and preserve the requested short/long architecture.
0019  5. **Convergence:** assign each ad a distinct territory, narrator cadence, mechanism expression, and payoff image.
0020  
0021  ## Truth-safe conversions
0022  
0023  - Fabricated experiment → explicit hypothetical planning device: `Try writing...`, `Picture...`, or `Use this question...`.
0024  - Composite first-person story → exact attributed VOC plus an ownership argument.
0025  - Invented dialogue → delete it or replace it with exact, source-mapped dialogue.
0026  - Unsupported result → desired absence, not a claimed customer outcome.
0027  - Unsupported timestamp/count → remove precision unless the value appears in the receipt or governing brief as a non-testimonial scenario.
0028  
0029  ## Deterministic revalidation
0030  
0031  For every repaired ad, calculate from the live Primary Text rather than trusting labels:
0032  
0033  - actual word count against the ad-specific range
0034  - required fields and exactly one finished ad
0035  - maximum two sentence endings per non-heading line
0036  - isolated dialogue
0037  - no em dash, banned words, stale price, or prohibited claims
0038  - exact offer token order across line breaks
0039  - exact CTA and destination
0040  - declared source VOC IDs
0041  - no-app/no-subscription consistency
0042  - mechanism-delivery compliance
0043  - exact quote presence for quoted VOC, with each excerpt verified as a literal source substring
0044  - explicit hypothetical framing where an artifact could read as a completed testimonial
0045  - no repeated substantive sentence across repaired ads, excluding mandated offer language
0046  
0047  Update the in-file verification count before hashing.
0048  
0049  Use this stable order after the final editorial change:
0050  1. Parse Primary Text and calculate actual counts.
0051  2. Update in-file count declarations.
0052  3. Re-run contract, banned-language, sentence-line, quote-fidelity, and cross-ad duplicate-line checks.
0053  4. Calculate final file, headline, and hook hashes.
0054  5. Rebuild the manifest from those final files.
0055  6. Re-run validation against the manifest.
0056  
0057  A count-label edit changes the file hash. Never calculate file hashes before count labels are final.
0058  
0059  Then regenerate the manifest from the final bytes, recording final word counts, checks, source IDs, and SHA-256 values.
0060  
0061  ## Final verification
0062  
0063  Re-open the manifest and all repaired ads, recompute hashes and counts, and fail delivery if any manifest value is stale.
0064  
0065  For source-sensitive repairs, add deterministic assertions for the actual blocker, such as:
0066  
0067  - exact approved quotation is present
0068  - known fabricated dialogue is absent
0069  - a `skip` mechanism ad contains no channel/placement explanation
0070  - a hypothetical log explicitly says it is hypothetical
0071  - an unsupported personal premise from the failed draft is absent
0072  
0073  Report only final status, files changed, material repair classes, and unresolved issues.

Native copy / references/control-format-selection-and-pilot-gate.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/control-format-selection-and-pilot-gate.md

SHA-256: a4c2ee41b3fd9a6e231a722b4d5ab70cbd93a2d014e3bcbd1f1c1db1d6b05926

Full source · 74 lines
0001  # Control-Format Selection and Long-Form Pilot Gate
0002  
0003  Use this when the user supplies example ads, says “like these samples,” asks for personal anecdotes, or requests many long-form natives.
0004  
0005  ## 1. Infer format from the controls before choosing a mode
0006  
0007  Measure the supplied controls rather than relying only on labels such as “native,” “copy,” or “variants.” Record:
0008  
0009  - primary-text word count
0010  - narrator type and point of view
0011  - whether the ad is scene-led or argument-led
0012  - product-name reveal percentage
0013  - opening object / incident
0014  - failed-solution depth
0015  - motif return and closing architecture
0016  
0017  Treat the samples as a format contract when their shape is unambiguous.
0018  
0019  ### Automatic Mode B signals
0020  
0021  Select DawnBands Long-Form Witness mode without re-asking when the request or controls clearly call for:
0022  
0023  - roughly 2,500–4,000-word executions
0024  - first-person personal anecdotes or witness stories
0025  - a lived world and progressive chronology before mechanism
0026  - product reveal around 55–70%
0027  - a concrete rupture object that returns at the close
0028  
0029  A request for “ten different angles” changes the story contracts, not the length or narrative architecture demonstrated by the supplied controls.
0030  
0031  Do not compress 3,000-word personal-story controls into 200–700-word ownership arguments merely because the user did not type the internal mode name.
0032  
0033  ## 2. Narrator selection
0034  
0035  When the user explicitly requests personal anecdotes and authorizes creative ad writing, `authorized-dramatization` may be the correct internal narrator mode if no verified buyer receipt exists.
0036  
0037  Keep behaviors grounded in avatar research and do not invent DawnBands outcomes, medical authority, quantified proof, or a named fake testimonial identity. The consumer-facing copy should still read as one coherent lived story. Never leak internal provenance policy into prose.
0038  
0039  Do not default to detached ownership voice when that would violate the sample’s first-person anecdotal format.
0040  
0041  ## 3. Pilot before batch scale
0042  
0043  For a batch of long-form ads:
0044  
0045  1. Build distinct story contracts for the whole batch.
0046  2. Draft one complete pilot with the requested model.
0047  3. Audit three windows, not just the hook:
0048     - opening 200–300 words
0049     - product reveal ±500 words
0050     - final 200–300 words
0051  4. Verify word count, reveal percentage, narrator intimacy, motif continuity, and lack of compliance/provenance leakage.
0052  5. Only then route the remaining lanes.
0053  
0054  A strong opening is not enough. Models often converge after the reveal into repeated phrases, identical mechanism lists, generic risk reversal, and cloned P.S. stacks.
0055  
0056  ## 4. Back-half diversity gate
0057  
0058  Across the finished batch compare:
0059  
0060  - first product-reveal sentence
0061  - bridge into product
0062  - mechanism analogy
0063  - objection order
0064  - offer phrasing
0065  - CTA architecture
0066  - final motif return
0067  
0068  If the lives differ but the reveals and closes share a template, preserve the story fronts and run same-model surgical rewrites from the first product mention through the end. Recount words and reveal percentages afterward.
0069  
0070  Useful deterministic signal: five-gram overlap can identify convergence, but a human read of opening, reveal, and close remains mandatory.
0071  
0072  ## 5. Delivery
0073  
0074  Preserve immutable raw model drafts, save same-model repairs separately, rerun deterministic QA, and publish a readable public report when the user expects to review thousands of words.

Native copy / references/controlled-live-meta-repair.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/controlled-live-meta-repair.md

SHA-256: 3b5a2dc98c372cb5cf456994c7432821961811bf45db9e42a910c09af6aad7de

Full source · 58 lines
0001  # Controlled Live-Meta Repair for a Brief-Driven Ad Batch
0002  
0003  Use this when repairing an existing native-ad batch after editorial QA and a current Meta evidence overlay identify a validated mechanism lane.
0004  
0005  ## Authority order
0006  
0007  1. User scope and file ceiling.
0008  2. Per-ad validated brief and current canonical product truth.
0009  3. Verified VOC/source receipts.
0010  4. Formal QA and sharpening blockers.
0011  5. Current live-Meta evidence as a prioritization overlay, not permission to override the brief.
0012  6. General copy doctrine and examples.
0013  
0014  ## Repair sequence
0015  
0016  1. Read each live ad immediately before editing.
0017  2. Consolidate all existing blockers and overlay deltas into one repair pass.
0018  3. Preserve each ad's persona, awareness, structure, word range, CTA, destination, and source IDs unless the brief explicitly permits a change.
0019  4. Convert unsupported first-person scenes, exact chronology, dialogue, and proof-shaped numbers into one of:
0020     - attributed exact evidence;
0021     - explicit hypothetical framing;
0022     - a non-testimonial ownership argument;
0023     - desired-absence framing.
0024  5. Allocate only the requested number of controlled evidence-backed lanes. Do not convert the whole batch into the current winner pattern.
0025  6. For two controlled ADHD biology-to-wrist lanes, keep this strategic sequence constant:
0026     - discipline/laziness shame relief;
0027     - concrete sourced alarm-stack recognition;
0028     - qualified biology/root-cause reframe;
0029     - why repeated sound cues fail;
0030     - direct wrist-vibration bridge;
0031     - parent-convenience or stop-being-the-alarm payoff;
0032     - exact risk reversal.
0033  7. Vary entry artifact, narrator mode, cadence, and mechanism phrasing between the controlled lanes. Declare the controlled lane and narrator mode in the ad metadata when auditability matters.
0034  8. Never clone a live winner's exact counts, timestamps, weekday, or signature phrasing. Live results validate a lane, not copy-only causality.
0035  9. Preserve short-form architecture. Do not inflate an Avelia T3 ad into long-form to satisfy a long-form SOP.
0036  10. For problem-aware long form, complete the sourced failed-attempt and root-cause chain before a short product reveal. Reveal percentage is diagnostic, not a universal threshold.
0037  11. Remove repeated product-description boilerplate unless a physical detail materially advances that ad's argument.
0038  12. Allocate exactly the number of controlled lanes the user requests. If the user names a lane letter or test-cell label, preserve it in the ad declaration and manifest; do not silently expand one requested lane into an older two- or three-lane default.
0039  13. When an exact source quote contains a globally banned weak word, do not alter text inside quotation marks. Use an exact contiguous excerpt that omits the weak word, or convert the line into an attributed paraphrase without quotation marks. Keep explicit attribution so the excerpt cannot read as advertiser testimony.
0040  14. Replace outcome-sounding setup statements such as `He agrees`, `He sets`, or `He wears` with capability or consent language such as `He has to agree`, `He can set`, or `He can wear` unless a receipt proves the completed behavior.
0041  
0042  ## Validation
0043  
0044  After every edit, parse the actual Primary Text section and verify:
0045  
0046  - brief-specific word range;
0047  - all required fields;
0048  - unique hooks and headlines within scope;
0049  - one sentence per line and no em dash;
0050  - banned language and claim restrictions;
0051  - exact CTA and destination;
0052  - canonical offer as three separate lines in exact order;
0053  - exact offer wording with no added prefix such as `DawnBands offers a`;
0054  - truth/provenance status for every quote, number, scene, timeline, and product outcome;
0055  - controlled lanes share the required strategy but differ in expression;
0056  - manifest SHA-256 values match the final files.
0057  
0058  Re-run validation after the manifest is written. Final status is PASS only when all expected files exist, all checks pass, and every hash matches.

Native copy / references/copy-directness-doctrine.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/copy-directness-doctrine.md

SHA-256: 726ba71b35a0dd28fe7e38a2562909579f72a481947331fa15f0374f01a89825

Full source · 51 lines
0001  # Copy Directness Doctrine (Fish, 2026-09-01) — BINDING ON ALL COPY
0002  
0003  Load with any native/ad copy task. This is the rule Fish enforces hardest and the one that
0004  has failed most often.
0005  
0006  ## Grade 3-5. Plain speech. Active voice.
0007  - Average sentence under 14 words. Most under 10. Nothing over 28.
0008  - Active voice only. "has been sorted into background by his brain" -> "his brain tuned it out."
0009  - Say the mechanism in ordinary words: the brain HABITUATES, it TUNES IT OUT, AUTOMATICALLY.
0010  
0011  ## The three-strike error: roundabout -> anaphora -> role-abstraction
0012  Fish rejected three consecutive rewrites of the same passage. Each fixed the surface and kept
0013  the literary move. Check for all three before shipping any line.
0014  
0015  1. ROUNDABOUT — "What I ordered was the first knock. ... the honest answer to what happens
0016     when he leaves home is that I am the answer."
0017  2. ANAPHORA (only sounds direct) — "To stop making the first knock. / To stop walking to his
0018     door. / To stop hearing my own voice before seven." Three+ consecutive lines opening with
0019     the same word or phrase is a poem. Banned.
0020  3. ROLE-ABSTRACTION — "right now I am still the plan." Fish: "i am still the plan is still
0021     not what a mom says." She never becomes an abstract noun.
0022  
0023  GOOD: "I am tired of being his alarm clock. I do not want to start every day angry at my own
0024  kid. He leaves next year, and I will still be the one calling him every morning."
0025  
0026  ## Banned constructions
0027  `I am still the plan` · `I am the answer` · `that was the variable` · `never in that channel` ·
0028  `that is the whole thing` · `that is what I am trying to fix` · `the part I could not fix was` ·
0029  `What I ordered was X` · `has been sorted`
0030  
0031  ## Instead
0032  Name the concrete action she performs, or the feeling she has.
0033  "I will still be the one calling him every morning." · "I still have to drive over and wake him
0034  up." · "I am tired." · "I feel guilty." · "I am angry before seven and I hate it."
0035  
0036  Register anchor, real buyers' words: "Want to start the school year off with out the fighting."
0037  "Hoping it stops angry mornings I wake up in fight or flight mode." "Mornings turned into daily
0038  fights."
0039  
0040  EXCEPTION: real idioms mothers use are fine — "I am his alarm clock", "the human snooze button".
0041  
0042  ## Test
0043  Would a 43-year-old mother say this sentence to another mother in a school parking lot?
0044  If it sounds written, cut it and say the feeling instead.
0045  
0046  ## Meta-lesson
0047  A fix that changes syntax but keeps the literary SHAPE is not a fix. Directness is not
0048  compression. It is plain speech.
0049  
0050  Enforced by `~/.openclaw/workspace/scripts/validate_native_copy.py`.
0051  Full history: `~/.claude/projects/-Users-ayden/memory/feedback_copy_directness_and_model.md`

Native copy / references/dawnbands-context-intake-and-precedence.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/dawnbands-context-intake-and-precedence.md

SHA-256: 05b3c75f619be609aec4058976104bddcc9bbb52a9eae759c54a60a26d4e1080

Full source · 54 lines
0001  # DawnBands native context intake and precedence
0002  
0003  Use this when reviewing, revising, or explaining an existing DawnBands native, especially when the user asks whether the full doctrine, briefs, avatar work, or research can be accessed.
0004  
0005  ## Intake rule
0006  
0007  Do not review from the rendered/public copy alone. Verify and read the internal source stack that produced it. Separate governing doctrine from historical examples and archives.
0008  
0009  ## Governing stack
0010  
0011  Read in this order, with later Fish rulings inside a higher source overriding older language:
0012  
0013  1. `native-copy/SKILL.md` — current permanent Fish rulings and execution gates.
0014  2. `references/long-format-native-sop-v2-contract.md` for long-format work.
0015  3. The current avatar/sub-avatar source for the unit. For the 2026-09 SA2 family: `knowledge/brands/dawnbands/research/evolve-subavatars-calmer-mornings-2026-09-03.md`.
0016  4. The unit's structured brief and system prompt in its output directory, including brief-gate receipts.
0017  5. Current brand truth: `_index.md`, `adhd-mom-lockin.md`, `dawnbands-avatar-adhd-teen-mom.md`, `product-truth.md`, `selling-root-cause-mechanism.md`, `substantiation-2026-08-06.md`, `failed-solutions-graveyard.md`, and `objection-map.md`.
0018  6. VOC receipts and current research that support the exact scene, qualifier, claim, and dialogue.
0019  7. Older research, swipe examples, generated drafts, and archives only where they do not conflict with items 1–6.
0020  
0021  ## Evolve work map
0022  
0023  For the calmer-mornings / Core 5 work, inspect all three together:
0024  
0025  - `knowledge/brands/dawnbands/research/inbox/fish-evolve-m1-notes-2026-09-02.md` — Fish's originating Evolve notes.
0026  - `knowledge/brands/dawnbands/research/evolve-subavatars-calmer-mornings-2026-09-03.md` — full SA1/SA2/SA3 build and rulings.
0027  - `knowledge/brands/dawnbands/research/winners-evolve-bins-2026-09-03.md` — performance classification and what remains untested.
0028  
0029  Do not treat the Evolve document's `DRAFT for Fish review` label as a reason to ignore rulings later embedded into the native-copy skill or product truth. Check whether a rule was promoted before deciding authority.
0030  
0031  ## Inciting-incident handling
0032  
0033  `knowledge/brands/dawnbands/research/inciting-incidents-2026-05-01.md` is useful research but was authored for the older broad mom-of-teen lane. It does not override the September ADHD-mom lock-in, a current Evolve sub-avatar card, the relatability gate, or a gated unit brief.
0034  
0035  For current long-form work, the binding standard is: write the incident at the peak of believability after the world, stack, and toll have established trust. The object needs a VOC receipt; the peak may be dramatized only in the sanctioned `incident_peak` slot and must stay confined to the narrator, child, and object.
0036  
0037  ## Existing-batch review
0038  
0039  When an output directory exists, inspect at minimum:
0040  
0041  - structured unit brief(s)
0042  - system/user generation prompts
0043  - immutable/raw draft
0044  - final copy
0045  - brief-gate audit
0046  - SOP audit
0047  - 1:1:1 receipt audit
0048  - launch/DCO plan if the question touches deployment
0049  
0050  A passing validator is not proof of editorial quality. Read the copy and the human judgment audit. Report conflicts or stale source assumptions rather than silently blending them.
0051  
0052  ## Public page boundary
0053  
0054  A Cloudflare Pages URL proves the rendered artifact is accessible. It does not prove the reviewer has loaded the doctrine, briefs, receipts, or source research. State which layer was actually inspected.

Native copy / references/dawnbands-long-form-witness-native.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/dawnbands-long-form-witness-native.md

SHA-256: d430ff955fdfc4637215f5310888062f6807072fda3c3ba707131cd93aa9bae5

Full source · 532 lines
0001  # DawnBands Long-Form Witness Native
0002  
0003  Use this when the brief explicitly selects **DawnBands Long-Form Witness Native** mode **or when the user's supplied controls clearly define that architecture**. A request to write “like these samples” is a format instruction when the controls are 2,500–4,000-word first-person witness stories with a late product reveal.
0004  
0005  It is an intentional long-form architecture derived from two supplied Alevia long-form controls, the Manus Long-Format Native Ad SOP 2.0, and the live DawnBands ADHD avatar/mechanism reports. It is not the default for every native, and it does not replace compressed Avelia T1/T2/T3 or product-aware ads.
0006  
0007  ## What the supplied controls prove structurally
0008  
0009  The two supplied Alevia controls were approximately 3,188 and 3,662 words. The product name first appeared at approximately 59.9% and 60.6% of the body. Both used the same persuasion engine:
0010  
0011  1. A witness sees a capable person suffer an identity rupture.
0012  2. One ordinary object holds the entire wound.
0013  3. The story establishes who that person was before the decline.
0014  4. Small losses accumulate before the inciting incident.
0015  5. Failed explanations and failed solutions are exhausted in sequence.
0016  6. Discovery happens through late-night research in the narrator's own voice.
0017  7. A plain-language mechanism makes prior failures feel inevitable.
0018  8. The product appears only after the reader can predict what a different solution must do.
0019  9. Results unfold through modest, concrete moments.
0020  10. The opening object returns at the end as proof of restored identity.
0021  11. Risk reversal removes the final objection.
0022  12. P.S. codas reopen or deepen emotional stakes rather than repeating features.
0023  
0024  Transfer the architecture, pacing, motif discipline, and emotional causality. Do not import unsupported medical claims, invented authorities, unverified timelines, or potentially fictional customer outcomes.
0025  
0026  ## Mode selector
0027  
0028  ### Use this mode when
0029  
0030  - The brief explicitly asks for an Alevia/Manus-style long-form witness story.
0031  - The user asks for personal anecdotes or says to match supplied 2,500–4,000-word first-person controls, even if the internal mode name is absent.
0032  - Traffic is unaware or problem-aware and needs story plus education.
0033  - There is enough avatar evidence to sustain a real world, not just a pain-point list.
0034  - The problem has a relational identity wound worth carrying for thousands of words.
0035  - The product can be introduced as the logical answer to a bounded mechanism.
0036  
0037  ### Do not use this mode when
0038  
0039  - Traffic is product-aware or offer-aware.
0040  - The brief requests compressed Avelia T1/T2/T3.
0041  - The only available source material is generic pain language.
0042  - A long story would require inventing a customer, authority, timeline, or DawnBands outcome.
0043  - The requested ad is a hook, static caption, testimonial card, or short objection killer.
0044  
0045  Product-aware readers get short proof, objection handling, offer, urgency, or risk reversal. Long-form is education, not a universal minimum.
0046  
0047  ## Length and reveal contract
0048  
0049  - Uploaded-winner-derived witness mode: typically **2,500–4,000 words** unless the brief sets another range.
0050  - Pure unaware Manus-style education may run longer when the world and mechanism genuinely earn it.
0051  - Product name target for witness mode: **55–70%** into the body.
0052  - Measured update 2026-08-12: Alevia's 239 live actives run reveal at median 52% (p10 31%,
0053    p90 66%) — the 55–70 contract stays default; **45–55% is a sanctioned test band** when the
0054    brief names it (`alevia-live-anatomy-2026-08-12.md`).
0055  - Pure unaware Manus mode may delay the product to 70–80%.
0056  - The brief and awareness level control. Never pad to hit a number.
0057  - Mechanism discovery may begin before the product reveal, but the category/product should not leak into the opening world.
0058  
0059  A 500-word ad can be strong, but it is not the same format as the supplied 3,000+ word controls. Do not label mid-form ownership copy as a full long-form witness native.
0060  
0061  ## Narrator authority gate
0062  
0063  Declare one narrator mode in the internal brief before drafting.
0064  
0065  ### 1. Verified buyer
0066  
0067  A single approved testimonial supports the narrator, product use, and claimed outcome.
0068  
0069  Allowed:
0070  - only details traceable to that source
0071  - exact or conservatively paraphrased outcomes
0072  - separately verified canonical product facts
0073  
0074  Not allowed:
0075  - adding a spouse, child, doctor, timeline, dialogue, or result the receipt does not contain
0076  
0077  ### 2. Single-source problem witness
0078  
0079  One first-person source supports the lived problem and witness perspective, but not DawnBands use.
0080  
0081  Allowed:
0082  - the sourced before-state and emotional wound
0083  - canonical DawnBands facts after the product bridge
0084  - separately attributed buyer proof
0085  - desired-absence future pacing
0086  
0087  Not allowed:
0088  - assigning DawnBands purchase, use, or results to the problem narrator
0089  
0090  ### 3. Ownership voice
0091  
0092  An ownership argument may reason from a sourced parent problem without pretending a pooled research synthesis is one mother's lived household testimony.
0093  
0094  Allowed:
0095  - attributed evidence from named source IDs
0096  - what a parent may fear, notice, or want to stop when framed as recognition rather than completed personal history
0097  - explicit hypotheticals and desired scenes
0098  - product facts and offer truth
0099  
0100  Not allowed:
0101  - a lived customer timeline or “what happened after we tried it” arc
0102  - welding multiple VOC sources into one `I / my son / our house` chronology
0103  - presenting a multi-device failed-solution stack, exact counts/times/distances, remembered dialogue, or a DIY research journey as one narrator's past unless one same-speaker receipt supports it
0104  - assuming an internal `ownership-voice` label cures consumer-facing testimonial implications
0105  
0106  **Synthetic-narrator tripwire:** if the prose combines first person, one child, one home, precise household behavior, and several failed solutions, stop. Use one same-speaker problem witness, attributed evidence, explicit wanted/feared hypotheticals, second-person recognition, or a non-testimonial ownership argument instead. A cleaner line edit must not make unsupported continuity feel more real.
0107  
0108  ### 4. Authorized dramatization
0109  
0110  Use only when the user or validated brief explicitly authorizes a dramatized/composite persona.
0111  
0112  - Label it internally.
0113  - Keep each behavior traceable to research.
0114  - Do not create a fake named testimonial identity.
0115  - Do not invent product use, product outcomes, week-by-week progression, medical authority, or quantified proof.
0116  - Any proof must be separately attributed to a real receipt.
0117  
0118  If none of the four modes is available, do not write first-person witness copy. Use attributed evidence, second-person recognition, or an ownership argument.
0119  
0120  ## Provenance ledger
0121  
0122  Tag every precise beat in the internal outline before prose:
0123  
0124  - `RECEIPT` — directly supported by one saved source.
0125  - `CANONICAL FACT` — supported by current DawnBands canonical truth.
0126  - `HYPOTHETICAL / OWNERSHIP ARGUMENT` — an explicitly framed wanted, feared, or likely scene that is not presented as completed history, product proof, or witness testimony.
0127  - `AUTHORIZED DRAMATIZATION` — a non-outcome connective detail allowed by the brief; never present it as customer proof or as a completed DawnBands result.
0128  - `DELETE` — unsupported precision, dialogue, authority, chronology, object, result, or claim.
0129  
0130  Specificity is proof-like. Names, ages, streets, schools, dates, exact times, counts, dollar amounts, objects, jobs, dialogue, and timelines require a receipt or explicit dramatization authority. “Plausible” is not provenance.
0131  
0132  ## DawnBands truth boundary
0133  
0134  The buyer and wearer are two different people sharing one morning problem.
0135  
0136  The bounded product role is:
0137  
0138  > A correctly fitted wrist alarm can provide a timed private tactile cue that is physically different from room-level sound.
0139  
0140  The product does not establish that cue detection becomes sustained waking, leaving bed, staying awake, punctuality, or reduced caregiver involvement for every wearer. Exact-SKU output, comfort, retention, battery behavior, reliability, and ADHD-specific efficacy remain test questions unless newer canonical evidence supersedes this file.
0141  
0142  Never claim DawnBands:
0143  
0144  - treats or fixes ADHD
0145  - fixes circadian timing or sleep inertia
0146  - resolves inadequate sleep, a sleep disorder, or medication questions
0147  - makes vibration impossible to sleep through
0148  - wakes every teen every time
0149  - guarantees sustained waking after cue detection
0150  - replaces medical evaluation for persistent sleep or health concerns
0151  
0152  Do not frame the teen as lazy, defiant, careless, or morally responsible for an experience they may barely remember. Preserve dignity and active wearer participation.
0153  
0154  ## Plain-language mechanism
0155  
0156  The defensible reframe is not “stronger alarm.”
0157  
0158  It is:
0159  
0160  - room sound broadcasts a cue through the environment
0161  - repeated versions of room sound can repeat the same failed pathway
0162  - wrist vibration changes both sensory channel and signal placement
0163  - a private cue can reduce whole-house spillover
0164  - the cue is only one handoff in the morning relay
0165  - alertness, initiation, leaving bed, staying awake, sleep opportunity, timing, and repeated-use habits still matter
0166  
0167  Preferred analogy:
0168  
0169  > The alarm is one handoff in a relay. DawnBands can move that handoff from the room to the wrist, but the rest of the morning still has to carry the baton forward.
0170  
0171  Vary the expression by ad. Do not clone “different sensory channel / different signal placement” across a batch.
0172  
0173  ## Story architecture
0174  
0175  ### Beat 1 — Rupture object
0176  
0177  Open inside a moment with one concrete object carrying the wound.
0178  
0179  DawnBands-compatible territories include:
0180  - untouched coffee cooling during another bedroom check
0181  - a backpack ready beside a teen who is still asleep
0182  - a phone ringing into an empty hallway
0183  - an unopened alarm box
0184  - a dorm packing pile that still assumes Mom will call
0185  
0186  The hook is not a headline, claim, benefit, diagnosis, or product clue. It is an event that makes line two necessary.
0187  
0188  Example objects are structural prompts, not receipts; the chosen object still needs `RECEIPT` or explicit `HYPOTHETICAL / OWNERSHIP ARGUMENT` framing.
0189  
0190  ### Beat 2 — Identity contrast
0191  
0192  Establish who the parent and teen are outside the problem.
0193  
0194  For the teen:
0195  - capable after waking
0196  - cares about school, work, sports, friends, or independence
0197  - does not need the ad to diagnose or excuse them
0198  
0199  For the parent:
0200  - patient before repetition changes their voice
0201  - trying to reduce support appropriately, not abandon the child
0202  - afraid the current arrangement will follow them into college or work
0203  
0204  The wound is not merely lateness. It is who each person becomes inside the current wake-up arrangement.
0205  
0206  ### Beat 3 — Progressive losses
0207  
0208  Show accumulation, not a generic pain summary.
0209  
0210  Possible source-backed progression:
0211  - one alarm becomes several
0212  - across-the-room becomes another unconscious shutoff
0213  - apps add tasks without sustained waking
0214  - Mom starts checking earlier
0215  - privacy turns into management
0216  - mornings begin with blame or guilt
0217  - future distance exposes the dependency
0218  
0219  Each loss must either have a receipt or be framed as a hypothetical/ownership argument.
0220  
0221  ### Beat 4 — Failed-solution sequence
0222  
0223  Use the order the avatar would logically try:
0224  
0225  1. louder ringtone
0226  2. more phone alarms
0227  3. phone across the room
0228  4. puzzle/QR/task app
0229  5. light or bed shaker
0230  6. repeated caregiver intervention
0231  
0232  For each:
0233  
0234  - acknowledge why it was logical
0235  - show its exact gap
0236  - give one memorable summary
0237  
0238  Do not insult prior choices. “Right instinct, wrong variable” is the tone.
0239  
0240  ### Beat 5 — Inciting incident
0241  
0242  The status quo becomes emotionally impossible to keep accepting.
0243  
0244  For DawnBands this does not need medical catastrophe. Stronger, truer incidents include:
0245  - the teen remembers only Mom's final angry attempt
0246  - a missed event mattered deeply to the teen
0247  - privacy is lost because Mom must enter and finish the morning
0248  - a college/work transition reveals that the system still depends on Mom
0249  - the parent realizes every alarm still ends with their body at the door
0250  
0251  The incident must be sourced or explicitly hypothetical. Never manufacture school calls, crashes, doctors, teachers, or dialogue for intensity.
0252  
0253  If the inciting pressure is not receipted, write it as a present-tense ownership realization or explicitly hypothetical threshold, never as a dramatized past event.
0254  
0255  ### Beat 6 — Discovery
0256  
0257  Two valid routes exist.
0258  
0259  #### Sourced authority route
0260  
0261  Use a real, traceable authority source. Paraphrase accurately. Do not fabricate a pediatrician, therapist, sleep researcher, pharmacist, or doctor conversation.
0262  
0263  #### DIY discovery route
0264  
0265  **DIY discovery** is valid when the narrator follows a real question into saved evidence.
0266  
0267  Structure:
0268  - the exact search question
0269  - the old assumption
0270  - the new distinction
0271  - a human-scale analogy
0272  - the moment prior failures make sense
0273  
0274  For DawnBands, the discovery is not “ADHD causes this.” It is the narrower realization that detecting a cue, becoming fully alert, and initiating the next action are different jobs, and that more room sound does not change signal placement.
0275  
0276  Measured refinement (2026-08-12): in every live Alevia survivor the discovery resolves as a
0277  **word spoken by a named third party inside the story** ("She told us about a fruit called…"),
0278  frequently at the end of a narrated chain — article → forum thread → a named stranger's comment
0279  → skeptical read → try. Narrating the chain itself does DR work. Two rules transfer: the
0280  mid-story reveal names the CATEGORY (a wrist-worn silent alarm someone mentions), never the
0281  brand — the brand surfaces at the link/close; and the named stranger stays dialogue inside the
0282  story, never an authority byline and never a fabricated testimonial with product outcomes.
0283  
0284  ### Beat 7 — Layered bounded mechanism
0285  
0286  Layer the explanation:
0287  
0288  1. What room sound asks the sleeper to notice.
0289  2. Why repetition or volume may still fail.
0290  3. Why task completion does not prove sustained waking.
0291  4. What changes when the cue is worn on the wrist.
0292  5. What remains unknown or outside the product's role.
0293  
0294  Stay in story voice. Use a reaction or concrete callback every 3–4 explanation lines so the copy does not become a strategy memo.
0295  
0296  ### Beat 8 — Objection sequence
0297  
0298  Handle objections in the narrator's natural order:
0299  
0300  - “He slept through a bed shaker. Why would vibration work?”
0301  - “Won't he get used to it?”
0302  - “He will not wear it.”
0303  - “I will end up managing another device.”
0304  - “What if this is really sleep deprivation, medication timing, or something medical?”
0305  - “Why risk $49 after all the other alarms?”
0306  
0307  Use acknowledge → gap → memorable summary. Do not promise impossible habituation resistance or universal efficacy.
0308  
0309  ### Beat 9 — Natural product reveal
0310  
0311  The product should feel like the answer the story has already specified.
0312  
0313  Keep the first reveal short and evidence-qualified:
0314  - seller-described silent vibrating wrist alarm for teens
0315  - cue intended to be worn directly on the wrist
0316  - standalone / no app / no subscription only after verifying the delivered revision
0317  - describe only visible form unless material and construction are independently verified
0318  
0319  Every exact-SKU statement must be classified internally as `VERIFIED`, `SELLER-CLAIMED`, or `UNKNOWN`. Consumer copy may state verified facts directly, attribute seller claims accurately when useful, and must omit unknowns.
0320  
0321  Do not switch into landing-page voice. Avoid “revolutionary,” “breakthrough,” “game-changing,” “proprietary,” and a sudden feature dump.
0322  
0323  ### Beat 10 — Proof or desired absence
0324  
0325  #### Receipt-backed outcome
0326  
0327  Use only when one approved DawnBands receipt supports the narrator, product use, and result. Preserve qualifiers and timeline exactly.
0328  
0329  #### Desired absence
0330  
0331  When narrator-level product proof is absent, future pace what the parent wants to stop:
0332  - no fourth walk down the hall
0333  - no whole-house alarm
0334  - no argument that begins before either person fully remembers the morning
0335  - no long-distance college wake-up call
0336  - no app that turns Mom into remote administrator
0337  
0338  Label the scene as wanted, imagined, or worth testing. Do not write it as a completed customer result.
0339  
0340  ## Outcome timeline gate
0341  
0342  No week-by-week, day-by-day, streak, “first morning,” teacher-noticed, spouse-noticed, or before/after progression may appear unless one traceable DawnBands user receipt supports that progression.
0343  
0344  If the gate fails:
0345  
0346  - remove the timeline
0347  - use separately attributed verified buyer proof
0348  - future pace as desired absence
0349  - move risk reversal forward
0350  
0351  The supplied Alevia controls use gradual timelines effectively. That structure is not permission to invent one for DawnBands.
0352  
0353  ## Motif contract
0354  
0355  Choose one concrete motif before drafting.
0356  
0357  The **motif** must:
0358  
0359  1. appear in the opening rupture
0360  2. reappear during mechanism/discovery as an analogy or reminder
0361  3. return in the close with changed meaning
0362  
0363  Examples:
0364  - cooling coffee: from interruption to a cup Mom gets to finish
0365  - bedroom door: from repeated crossing to a door she does not open
0366  - backpack: from readiness blocked by waking to readiness owned by the teen
0367  - unopened box: from category fatigue to a bounded different-cue test
0368  
0369  One motif only. Do not add a second cinematic symbol because the first one was not developed.
0370  
0371  ## Close and P.S. architecture
0372  
0373  Close in this order:
0374  
0375  1. return to the opening motif
0376  2. restate the relational loss of doing nothing
0377  3. make the bounded product role clear
0378  4. state only the currently verified price and refund-policy terms
0379  5. one direct CTA
0380  
0381  Offer and guarantee boundary:
0382  
0383  - Verify the current price and complete return/refund policy before every deployment.
0384  - Prefer unambiguous policy language such as `60-night money-back guarantee` only when those terms are verified.
0385  - `100% wake-up guarantee` may appear only if legal/substantiation review has approved it as the exact name of a refund policy, and it must be immediately clarified as refund terms rather than universal waking efficacy.
0386  - Never write the guarantee as proof that DawnBands wakes every wearer, sustains waking, creates independence, or produces a guaranteed outcome.
0387  - If current policy evidence is incomplete or conflicting, stop the offer block and request verification rather than inheriting older canonical copy.
0388  
0389  Optional codas:
0390  - P.S. returns to the most emotional unresolved relationship beat.
0391  - P.P.S. handles the dominant risk or skepticism objection.
0392  - P.P.P.S. may extend the future/college implication when genuinely earned.
0393  
0394  Do not use codas as repeated feature bullets, fake scarcity, or unsupported stock claims.
0395  
0396  ### Close variant B — the measured P.S. ladder (added 2026-08-12)
0397  
0398  217 of 240 measured live Alevia actives close on a stacked coda ladder instead of the ordered
0399  close above, with the link LAST. Test it against the default close on a proven copy before
0400  adopting it batch-wide (see `alevia-live-anatomy-2026-08-12.md`):
0401  
0402  1. P.S. — rupture-object/motif return
0403  2. P.P.S. — guarantee naturalized in the narrator's voice (verified refund terms only; every
0404     offer gate above applies)
0405  3. P.P.P.S. — the dominant objection or "it's just placebo" kill, in-voice
0406  4. final P.S. — direct address to the reader-avatar; this rung IS the CTA
0407  5. bare 👉 link on its own line after the final coda — no CTA copy around it
0408  
0409  Outcome-timeline, desired-absence, and offer-verification gates apply to every rung. The ladder
0410  never becomes a feature recap.
0411  
0412  ## Formatting
0413  
0414  **(Corrected 2026-08-12 per the 2026-07-31 Mode B ruling + measured live controls — see
0415  `alevia-live-anatomy-2026-08-12.md`. The older one-sentence-per-line block that stood here
0416  applies to Mode A / compressed natives and static primaries, NOT to this mode.)**
0417  
0418  - **PARAGRAPH prose** — 2–4 sentences per paragraph, roughly 25–40 words. The measured live
0419    controls run ~31 words/paragraph median. A witness story formatted wall-to-wall
0420    one-sentence-per-line reads as an ad, not a post.
0421  - Staccato single-line beats are a PUNCH device, not the default — reserve them for rupture
0422    moments, reveals, and emotional pivots (the live controls mix them into roughly half the
0423    copies, sparingly).
0424  - blank line between paragraphs
0425  - short, varied cadence within paragraphs
0426  - dialogue isolated on its own line
0427  - no emojis in long-form story copy (the bare 👉 link line at the close is the one exception)
0428  - no bold/caps/format tricks inside Meta primary text
0429  - no em dash
0430  - no dense mechanism wall
0431  - no visible switch into ad voice
0432  
0433  ## Internal pre-draft contract
0434  
0435  ```yaml
0436  mode: dawnbands-long-form-witness
0437  submode: alevia-witness | pure-unaware-manus
0438  awareness: unaware | problem-aware
0439  word_target: 2500-4000
0440  narrator_mode: verified-buyer | single-source-problem-witness | ownership-voice | authorized-dramatization
0441  primary_source_ids: []
0442  product_outcome_receipt: null
0443  rupture_object: ""
0444  identity_before: []
0445  progressive_losses: []
0446  failed_solution_sequence: []
0447  inciting_incident:
0448    beat: ""
0449    provenance: RECEIPT | HYPOTHETICAL_OWNERSHIP
0450  mechanism_route: sourced-authority | diy-discovery
0451  primary_open_loop: ""
0452  secondary_loops: []  # require at least 1 in Mode B unless the brief explicitly waives it
0453  product_reveal_target_pct: 55-70
0454  motif_return: ""
0455  future_mode: receipt-backed-outcome | desired-absence
0456  canonical_truth_sources: []
0457  exact_sku_claims:
0458    verified: []
0459    seller_claimed: []
0460    unknown_delete: []
0461  offer_policy_verified: false
0462  wake_up_guarantee_legal_approval: false
0463  claim_ceiling_checked: false
0464  ```
0465  
0466  Stop if submode, narrator authority, source IDs, mechanism route, inciting-incident provenance, product outcome receipt/null decision, primary open loop, at least one secondary loop or explicit waiver, motif, future mode, exact-SKU claim classification, verified current offer/refund evidence, or claim ceiling is missing.
0467  
0468  ## QA scorecard
0469  
0470  ### Architecture
0471  
0472  - [ ] Intentional witness mode was selected by the brief.
0473  - [ ] Hook is a story event with no product/category leak.
0474  - [ ] Identity contrast appears before mechanism.
0475  - [ ] Progressive losses accumulate.
0476  - [ ] Failed solutions follow a believable sequence.
0477  - [ ] Inciting incident is irreversible enough and provenance-safe.
0478  - [ ] One primary open loop is introduced in the opening world and stays unanswered until after mechanism/discovery.
0479  - [ ] At least one secondary loop deepens stakes, proof, or future risk before the reveal.
0480  - [ ] Every loop is closed in the body or intentionally handed to a P.S. / P.P.S. coda.
0481  - [ ] Product reveal lands in the selected window.
0482  - [ ] Reveal stays in narrator voice.
0483  - [ ] Motif returns with changed meaning.
0484  
0485  ### Avatar
0486  
0487  - [ ] Buyer and wearer remain distinct people.
0488  - [ ] Teen dignity and agency are preserved.
0489  - [ ] The parent is not framed as controlling for its own sake.
0490  - [ ] Relationship cost is shown through behavior.
0491  - [ ] College/work future is used only when the angle earns it.
0492  
0493  ### Truth and proof
0494  
0495  - [ ] Every precise detail has a provenance label.
0496  - [ ] No composite becomes a fake testimonial.
0497  - [ ] No pooled VOC set is presented as one `I / my son / our house` chronology without a same-speaker receipt or explicit dramatization authority.
0498  - [ ] First-person specificity, failed-device history, counts/times/distances, and discovery behavior pass the synthetic-narrator tripwire.
0499  - [ ] Quoted words match the saved source.
0500  - [ ] No authority dialogue was invented.
0501  - [ ] No product outcome or timeline exceeds its receipt.
0502  - [ ] Desired scenes are not written as completed results.
0503  - [ ] Desired-absence scenes are framed with wanted / imagined / worth-testing language, never as completed DawnBands results.
0504  - [ ] Exact-SKU statements are classified as verified, seller-claimed, or unknown; unknowns are omitted.
0505  - [ ] Current price and complete refund-policy terms are verified before the offer block.
0506  - [ ] `100% wake-up guarantee` is absent unless approved as a named refund policy and immediately clarified as refund terms, not universal waking efficacy.
0507  - [ ] Exact-SKU uncertainty and cue-vs-waking boundary are preserved.
0508  
0509  ### Craft
0510  
0511  - [ ] One concrete motif, not several.
0512  - [ ] DIY discovery or authority explanation sounds human, not like a paper.
0513  - [ ] Each failed solution gets acknowledge → gap → summary.
0514  - [ ] Mechanism explanation returns to the lived story every 3–4 lines.
0515  - [ ] CTA resolves the opening tradeoff.
0516  - [ ] Optional P.S. codas deepen the story rather than repeat sales copy.
0517  
0518  ## Batch rule
0519  
0520  Do not make every ad in a batch a 3,000-word witness story. Long-form witness should be an intentional testing cell alongside compressed problem-aware, solution-aware, product-aware, and objection-led executions.
0521  
0522  Within a long-form cell, vary:
0523  - narrator mode
0524  - rupture object
0525  - primary fear
0526  - failed-solution stack
0527  - mechanism analogy
0528  - product reveal route
0529  - motif
0530  - desired absence
0531  
0532  The architecture may repeat. The life cannot.

Native copy / references/dawnbands-selling-root-cause-mechanism.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/dawnbands-selling-root-cause-mechanism.md

SHA-256: abe4e68240dfa2edfdf1e9b30cb7f1826c33997993caa0d25c6c81d8d00d4b8c

Full source · 88 lines
0001  # DawnBands Selling Root Cause + Mechanism — Execution Reference
0002  
0003  Use this reference for any DawnBands native, advertorial, UGC, VSL, listicle, lander, brief, or copy audit.
0004  
0005  The workspace canon remains authoritative:
0006  
0007  `/Users/ayden/.openclaw/workspace/knowledge/brands/dawnbands/selling-root-cause-mechanism.md`
0008  
0009  ## User correction captured
0010  
0011  Do not over-optimize DawnBands mechanism copy for conservative addressability. Aggressive biological simplification is desired when it makes the Root Cause → Mechanism → Product chain more persuasive.
0012  
0013  Keep two layers separate:
0014  
0015  - **Selling layer:** exaggerated, simplified biological explanation that creates belief and new hope.
0016  - **Proof layer:** receipts for exact-SKU performance, testimonials, timelines, offer terms, and customer outcomes.
0017  
0018  The proof layer must not flatten the selling mechanism into generic feature language. The selling layer does not authorize fabricated customer history or product proof.
0019  
0020  ## Canonical selling chain
0021  
0022  `circadian delay → sleep debt → auditory gating → sleep inertia / half-awake automatic behavior → Mom becomes the emergency pathway → Somatic Wake Bypass™ → Wrist-First Handoff™ → relationship repair + future independence`
0023  
0024  ## Root cause: The Deep-Sleep Signal Lock
0025  
0026  1. **Circadian delay:** His teenage body clock is still protecting biological night when school demands consciousness.
0027  2. **Sleep debt:** Accumulated sleep pressure is pulling harder toward sleep while the alarm pushes toward waking.
0028  3. **Auditory gating:** Repeated room sound is suppressed or pattern-matched as background.
0029  4. **Sleep inertia:** His body can silence alarms, answer Mom, or solve tasks before alertness, memory, and initiation fully return.
0030  5. **Emergency pathway:** Mom escalates until emotion becomes the only novel signal left.
0031  
0032  Core lines:
0033  
0034  - “The alarm wakes his hands without waking the part of him that can start the morning.”
0035  - “The louder the room gets, the better his sleeping brain becomes at deleting it.”
0036  - “His brain ignores the alarms, but it notices the anger.”
0037  - “She remembers nine patient attempts. He remembers waking up to one angry mother.”
0038  
0039  ## Failed-solution trap
0040  
0041  > Seven alarms do not create seven new chances to wake up. They create seven repetitions of the same signal his brain already filed under background noise.
0042  
0043  Household loop:
0044  
0045  > The more reliably Mom rescues the failed alarm, the more reliably the morning keeps depending on her.
0046  
0047  Never blame Mom. The villain is the repeated room-sound rescue system that leaves her no safe way to step out.
0048  
0049  ## Biological mechanism: Somatic Wake Bypass™
0050  
0051  DawnBands puts a tactile signal against the wrist, outside the repeated room-sound pattern.
0052  
0053  Core line:
0054  
0055  > Sound has to cross the room and survive his brain’s filter. DawnBands starts past the filter.
0056  
0057  ## Household mechanism: Wrist-First Handoff™
0058  
0059  The first cue begins on the teen before Mom’s voice enters the morning.
0060  
0061  Core lines:
0062  
0063  - “The first cue starts on his wrist before it ends in your voice.”
0064  - “Not louder. On him.”
0065  - “Move the first signal from your throat to his wrist.”
0066  - “The goal is not a better tenth attempt. It is a first cue that does not require Mom.”
0067  
0068  ## Format routing
0069  
0070  - **Long problem-aware:** use the complete biological and household chain.
0071  - **Short native:** compress to auditory gating → tactile bypass → Mom removed from first cue.
0072  - **UGC:** explain that the body can perform an alarm-stopping action before memory and conscious initiation are online.
0073  - **Advertorial/VSL:** name Deep-Sleep Signal Lock and Somatic Wake Bypass™ after lived pain and failed solutions.
0074  - **Product-aware:** “Sound has to cross the room and survive his brain’s filter. DawnBands starts past the filter.”
0075  
0076  ## Audit gate
0077  
0078  Where format length permits, future DawnBands copy should answer:
0079  
0080  1. Why is his biology pulling him toward sleep?
0081  2. Why does repeated sound fail?
0082  3. Why can he act without becoming consciously awake?
0083  4. Why does Mom become the final signal?
0084  5. Why is tactile wrist stimulation a new pathway?
0085  6. Why does the first cue now belong to him?
0086  7. Why does this matter to their relationship and his future?
0087  
0088  If a draft only says “vibration is different from sound,” the mechanism is underbuilt.

Native copy / references/direct-failed-solution-prose.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/direct-failed-solution-prose.md

SHA-256: 87b28977039d10e41f9f37ec62e4e658898b67e6c16c52da353254d862059f74

Full source · 48 lines
0001  # Direct Failed-Solution Prose
0002  
0003  Use this when a native contains a sequence of things the narrator tried before discovering the product.
0004  
0005  ## The unit
0006  
0007  Each attempt gets one complete causal beat:
0008  
0009  1. **What she tried** — literal action, not a method label.
0010  2. **Why she tried it** — the belief or hope behind it.
0011  3. **What happened** — visible behavior in the morning.
0012  4. **Why trust broke** — the same problem still occurred.
0013  
0014  Example:
0015  
0016  > I started getting up at 5:30 because I thought I was the problem. I wanted to be ready and calm before he got up. I still ended up yelling by the door.
0017  
0018  ## Sequence rules
0019  
0020  - One attempt per beat. Keep chronology obvious.
0021  - Describe actions before using specialist labels.
0022  - Narrator-specific attempts are not reader qualifiers.
0023  - Connect every attempt to the owned morning problem; do not list generic parenting advice.
0024  - State the failure directly. Do not make the reader infer it from a slogan.
0025  - If two tactics are closely related, distinguish what changed and what remained under Mom's control.
0026  
0027  ## Kill on sight
0028  
0029  - `I tried everything the posts say`
0030  - `came next`
0031  - `laid it on`
0032  - `used it up`
0033  - `both helped`
0034  - `neither stopped`
0035  - several tactic names compressed into fragments
0036  - unexplained before/after jumps
0037  - a short sentence that still needs translation
0038  
0039  ## Review test
0040  
0041  For every attempt, answer in one read:
0042  
0043  - What did she physically do?
0044  - Why did she think it would help?
0045  - What did her son physically do?
0046  - Why did she stop believing this would fix the morning?
0047  
0048  If any answer is missing, rewrite the beat. Numeric readability is not enough: grade, sentence length, and said-out-loud meaning must all pass.

Native copy / references/directness-and-jumble-final-pass.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/directness-and-jumble-final-pass.md

SHA-256: cb65ddf502af44156ddeffc70a399c110afa235edd4d01859b646cb83fa27d68

Full source · 63 lines
0001  # Directness and Final Jumble Pass
0002  
0003  Use after structural/SOP validation on any long native. Mechanical readability can pass while a paragraph still makes a reader stop and decode it.
0004  
0005  ## Core distinction
0006  
0007  - **Readable numbers:** grade, sentence length, paragraph length.
0008  - **Readable meaning:** a normal reader understands the action, chronology, speaker, and cause on the first pass.
0009  
0010  Both must pass. A short sentence can still be jumbled.
0011  
0012  ## Failed-solution unit
0013  
0014  Write one attempt per beat in this order:
0015  
0016  1. what she tried
0017  2. why she tried it
0018  3. what literally happened
0019  4. why she stopped trusting it
0020  
0021  Do not compress two methods into one summary paragraph. Avoid transition shorthand such as `came next`, `both helped`, `used it up`, or `everything the posts say`. Name the action and result.
0022  
0023  **Direct model:**
0024  
0025  > I started getting up earlier because I thought I was the problem. I wanted to be ready and calm before he got up. I still ended up yelling by the door.
0026  
0027  ## Paragraph-level jumble checklist
0028  
0029  Read every paragraph and flag:
0030  
0031  - two speakers whose turns are hard to distinguish
0032  - an unclear `it`, `that`, `under it`, `both`, or `they`
0033  - a chronology jump without a transition
0034  - multiple failed solutions compressed into one beat
0035  - framework labels leaking into the narrator's words
0036  - abstract status descriptions instead of visible behavior
0037  - cause/effect that skips a step
0038  - a fragment whose meaning depends on reconstructing the previous paragraph
0039  - product/mechanism phrasing that sounds like an operator note
0040  - a callback whose original object or event is no longer clear
0041  
0042  ## Common transformations
0043  
0044  - `He was just not started` → show the stalled action: `He stood beside his shoes without reaching for them.`
0045  - `He can hear you. He cannot begin.` → `He hears you, but his body still does not move.`
0046  - `Your calm is his calm` → state what each person sees/hears and how tension increases.
0047  - `I got myself out of the wake-up` → `I stopped being the person who had to wake him.`
0048  - `The first thing he does is his own` → `You do not have to give him the first order.`
0049  - `He comes up with a clearer head` → `He wakes with a clearer head.`
0050  - `Gets lost between fewer steps` → `Loses track of fewer steps.`
0051  - `See what the rest of the hour does` → state the observable result being tested.
0052  
0053  ## Final sequence
0054  
0055  1. Run the governing native validator and SOP mechanical audit.
0056  2. Read the full body cold, paragraph by paragraph.
0057  3. Apply only line-level directness repairs; preserve the approved structure and incident.
0058  4. Rerun sentence-length, grade, banned-language, and cross-body convergence checks.
0059  5. Complete the SOP §11 eight-killers table and §13 ten-question final check.
0060  6. Verify the receipt table still passes after wording changes.
0061  7. Rebuild the review artifact only after the final text is frozen.
0062  
0063  A deterministic PASS is not permission to skip this pass.

Native copy / references/explicit-batch-cardinality-and-publication.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/explicit-batch-cardinality-and-publication.md

SHA-256: 89b4cb6564039cc0e79859b5add1c1abada467732085ff4b68e32fc9009283fe

Full source · 39 lines
0001  # Explicit Native Batch Cardinality and Publication Boundary
0002  
0003  Use when Fish names the exact count of bodies, images, hook variants, headlines, or ads, or says to “publish” a completed native batch.
0004  
0005  ## Cardinality math
0006  
0007  Treat the user's explicit layout as the contract even when it differs from the standing test default.
0008  
0009  - One **body** is one message/architecture.
0010  - One **body × image pair** is one ad unit unless Fish explicitly defines another counting rule.
0011  - Hook variants and headline variants may ride inside Meta's multi-text fields and do **not** multiply the ad-unit count by themselves.
0012  - Example: 3 bodies × 3 images = 9 ads. If each ad carries 2 primary-text hook variants and 3 headlines, the manifest contains 9 ad IDs, 18 primary-text slots, 27 headline slots, and 9 unique image assignments.
0013  
0014  Before delivery, assert the requested arithmetic mechanically and record unique ad IDs, unique image paths, primary-text slots per ad, headline slots per ad, destination, and CTA. Do not silently fall back to the standing 3-hook/5-image layout after Fish gives exact numbers.
0015  
0016  ## Reader qualifiers versus narrator texture
0017  
0018  The relatability gate is strictest in reader-identification fields (`hook`, `control_hook`, `stake`). Age, grade, transport, work details, and household scene details may be legal narrator texture when they stay attached to this narrator, carry a VOC/narrator/sanctioned-dramatization receipt, and do not create a second sub-avatar.
0019  
0020  Do not ban a receipted missed bus or teen age from a narrator's incident merely because it would be an unverified reader qualifier. Keep siblings and partners out unless the card/receipt authorizes them.
0021  
0022  ## Reviewer adjudication
0023  
0024  A challenger is an input, not the final source of truth.
0025  
0026  - Accept genuine science, causality, claim-ceiling, or provenance corrections.
0027  - Retire a challenger finding when it conflicts with newer canonical product truth or an explicit Fish-approved dramatization contract; write the reason in the human audit.
0028  - Authorized dramatization remains internally labeled and must never be represented as a verified customer testimonial.
0029  - Controlled SOP constants may repeat. Check unsanctioned convergence separately with exact-line and five-gram scans after excluding those constants.
0030  
0031  ## “Publish” boundary
0032  
0033  If Fish says “publish” but supplies no Meta campaign/ad-set target, budget, bid, or launch destination, publish and verify the public review artifact and produce the complete launch manifest, but do not infer a Meta launch target.
0034  
0035  A Meta launch still needs a resolvable campaign/ad-set destination and all applicable explicit approvals. Distinguish “review page published” from “Meta launched” in the completion note.
0036  
0037  ## Verification receipt
0038  
0039  For public review delivery, verify both the stable Pages URL and immutable deployment URL return HTTP 200 and contain markers for every body/ad ID. Probe representative image assets directly as well.

Native copy / references/lived-first-person-native-copy.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/lived-first-person-native-copy.md

SHA-256: 50dc4c9ebf42a0f7d4cb5759996915d5bbe2d80e757239090894ea670d6f8c60

Full source · 62 lines
0001  # Lived First-Person Native Copy Without Fake Testimonial Claims
0002  
0003  Use this when a native ad is structurally correct but sounds like an analyst describing the avatar from outside.
0004  
0005  ## The failure
0006  
0007  Detached copy talks about:
0008  - “Mom,” “the buyer,” “the category,” or “the sleeper”
0009  - solution architecture before a lived trigger
0010  - generalized emotional labels instead of actions
0011  - a product result assigned to a narrator who was never verified
0012  
0013  This can satisfy a brief while missing the essence of native: the reader should feel the writer is inside the morning, not presenting a market analysis.
0014  
0015  ## Lived-POV construction
0016  
0017  1. Open with one observable trigger and one consequence.
0018  2. Stay in first-person logistics: what I saw, did, repeated, heard, or regretted.
0019  3. Use short sequence lines to recreate the loop rather than summarize it.
0020  4. Name the identity wound in the narrator's language only after the scene earns it.
0021  5. Correct the failed category after the reader recognizes herself.
0022  6. Reveal the product as a different test, not as a sudden sales pitch.
0023  7. Put verified customer outcome proof in a separately attributed quote.
0024  8. Close by resolving the narrator's role or tradeoff.
0025  
0026  ## Truth-safe first-person modes
0027  
0028  Use `I` only when at least one mode is explicit:
0029  
0030  - **Verified buyer:** exact review/testimonial supports the narrator and product outcome.
0031  - **Single-source problem POV:** one first-person VOC source supports the lived problem; do not assign product use or outcome to that narrator.
0032  - **Authorized dramatization/composite:** the brief or user explicitly authorizes a dramatized persona voice. Label this internally, keep every behavior receipt-backed, and never invent a product result, timeline, or testimonial identity.
0033  - **Ownership voice:** first-person problem language transitions to brand/product facts, with any result proof separately attributed.
0034  
0035  If none applies, use attributed VOC or second-person recognition rather than manufacturing a customer.
0036  
0037  ## Detached-language gate
0038  
0039  Fail and rewrite if the body could naturally contain phrases such as:
0040  - “another morning Mom has to believe in”
0041  - “that skeptical buyer”
0042  - “the category keeps changing its packaging”
0043  - “the hoped-for result”
0044  
0045  Those may be useful strategy notes, but they are not how a mom posts in a parenting group.
0046  
0047  ## Avelia transfer rule
0048  
0049  Borrow Avelia's causal compression, not its surface voice:
0050  
0051  `one trigger → concrete symptom/behavior → failed explanation or category → different mechanism → compressed proof/risk reversal → resolve the opening tradeoff`
0052  
0053  For first-person parent native, the trigger must be lived before the mechanism is explained.
0054  
0055  ## Quick QA
0056  
0057  - Is line 1 something the narrator personally saw/did, not an analyst's thesis?
0058  - Do lines 3–8 contain concrete sequence detail?
0059  - Does the product appear only after the lived problem is established?
0060  - Is every customer outcome attached to a real, named/attributed receipt?
0061  - Could this plausibly be posted by the target person without sounding like a copywriter?
0062  - Does the close resolve “who I had to become” or “what I had to keep doing”?

Native copy / references/long-form-native-comparative-audit.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/long-form-native-comparative-audit.md

SHA-256: 5ab65a462a7697a06d9ddb6d88acaa613d3606d8b89dbfe2ae20aa87a9a181dc

Full source · 110 lines
0001  # Long-Form Native Comparative Audit
0002  
0003  Use this for a read-only audit of a native-ad batch against competitor/winner copies, a live long-form SOP, and live avatar/mechanism reports.
0004  
0005  ## Authority and evidence order
0006  
0007  1. The newest live product/mechanism report governs the claim ceiling.
0008  2. Verified exact-SKU facts and current policy terms govern product and offer language.
0009  3. Live avatar/VOC reports govern audience scenes, objections, and language, but problem VOC is not product proof.
0010  4. The live SOP governs architecture by awareness level.
0011  5. Winner copies supply structural hypotheses only. Their narrator, dialogue, precision, medical claims, results, scarcity, and performance status do not transfer.
0012  6. Older brand doctrine, gold standards, and briefs are subordinate when they conflict with a newer live evidence report. Record the conflict explicitly rather than silently choosing one.
0013  
0014  ## Required distinctions
0015  
0016  Report separately:
0017  
0018  - **Structural quality:** hook, world, accumulation, failed attempts, inciting incident, discovery, mechanism, objections, reveal, resolution, close, and formatting.
0019  - **Truth/provenance:** narrator identity, quotation fidelity, exact times/counts, dialogue, outcomes, authority, exact-SKU facts, medical/performance claims, and offer policy.
0020  - **Format fit:** full long-form, medium mechanism story, or compressed/product-aware execution. Do not penalize a valid compressed ad for omitting a 500-line story.
0021  
0022  A structurally strong ad can fail provenance. A truth-safe ad can still fail long-form architecture.
0023  
0024  ## Deterministic audit pass
0025  
0026  For every ad, calculate from the actual Primary Text:
0027  
0028  - body word count;
0029  - first product-name line;
0030  - first product-name word position as a percentage of body words;
0031  - awareness level and intended format;
0032  - repeated product/mechanism sentences across the batch.
0033  
0034  Use those measurements as evidence, not as automatic grades. For intentional unaware/problem-aware long form, compare reveal timing with the governing SOP. Early reveal can be correct for compressed or product-aware ads.
0035  
0036  ## Device matrix
0037  
0038  Create one row per winner/SOP device:
0039  
0040  - story-beginning hook;
0041  - main open loop;
0042  - avatar world and before-state;
0043  - gradual accumulation;
0044  - relational stakes;
0045  - failed-solutions sequence;
0046  - irreversible inciting incident;
0047  - credible discovery or authority bridge;
0048  - layered mechanism with human processing beats;
0049  - analogy;
0050  - objection handling disguised as story;
0051  - late invisible product reveal;
0052  - modest future pacing or exact sourced results;
0053  - relationship resolution;
0054  - victory as absence;
0055  - loss-aware close and verified risk reversal;
0056  - callback/P.S. architecture;
0057  - native formatting and real-person voice.
0058  
0059  Grade each `present`, `partial`, `absent`, or `misused`. Cite exact ad IDs, line numbers, and short excerpts. Explain whether the issue is structural, evidential, or both.
0060  
0061  ## Per-ad ledger
0062  
0063  Give every expected ad one row containing:
0064  
0065  - format/awareness fit;
0066  - strongest owned device or artifact;
0067  - principal structural weakness;
0068  - principal provenance/claim risk;
0069  - confidence/blocker status.
0070  
0071  Do not let a batch-level summary hide an individual ad.
0072  
0073  ## Composite-testimonial rule
0074  
0075  Metadata such as “dramatized,” “composite,” or “grounded in VOC” does not cure consumer-facing first-person testimony. If the body says `I`, `my son`, or presents a chronology as something that happened, require one same-speaker receipt for the narrator, event, dialogue, precision, and result.
0076  
0077  Truth-safe alternatives:
0078  
0079  - exact attributed VOC;
0080  - explicit hypothetical language in the consumer copy;
0081  - a non-testimonial ownership argument;
0082  - desired-absence framing;
0083  - an exact-SKU customer receipt with preserved attribution.
0084  
0085  ## Winner-copy boundary
0086  
0087  Extract devices from a winner, but audit the winner itself for proof-shaped fabrication. Named people, doctors, locations, clinical numbers, week-by-week outcomes, authority dialogue, and scarcity remain unusable without receipts. Do not describe longevity, duplicate count, or labels such as “winner” as causal performance proof unless a closed performance window supports it.
0088  
0089  ## Doctrine output
0090  
0091  End with a locked doctrine specific to the brand that defines:
0092  
0093  1. when full long-form is allowed by awareness;
0094  2. the evidence hierarchy;
0095  3. the required story sequence;
0096  4. allowed, qualified, and prohibited mechanism language;
0097  5. reveal timing by format;
0098  6. proof/offer verification gates;
0099  7. cross-ad uniqueness requirements;
0100  8. the rule that QA language belongs in audit notes, not consumer-facing prose.
0101  
0102  ## Read-only verification
0103  
0104  Before delivery confirm:
0105  
0106  - all expected ads were audited;
0107  - all sub-7 or blocker findings have exact evidence and, when requested, proposed text;
0108  - structural weaknesses are separated from truth/provenance risks;
0109  - source files were not edited;
0110  - pre-existing repository changes are not attributed to the audit.

Native copy / references/long-format-native-sop-v2-contract.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/long-format-native-sop-v2-contract.md

SHA-256: e4d2bee55a7b7a9f54d57aa15c066b034851ea4d166579da0eb5675ce29313b8

Full source · 123 lines
0001  # Long-Format Native Ad SOP 2.0 — binding contract (Fish 2026-09-03: "follow the doc and what it says. It is more true.")
0002  
0003  Source of truth: Zakaria, "Long-Format Native Ad SOP 2.0" (https://native-ads.manus.space/), full text at
0004  `~/.openclaw/workspace/knowledge/formats/zakaria-long-format-native-sop-v2-2026-09-03.txt`. This file turns it
0005  into the rules every long-format native is written and validated against. Where an older reference in this
0006  skill disagrees (Alevia-derived witness contract, T1/T2 budgets, victory-as-absence, reveal window,
0007  paragraph prose, analogy ban inside the mechanism), THIS FILE WINS. Audit of the deltas:
0008  `knowledge/formats/native-sop-v2-audit-2026-09-03.md`.
0009  
0010  One fence stays because the SOP itself requires it ("true or defensibly true", "the reader will Google it"):
0011  never invented science; the mechanism is the ruled chain and the one verified study. The 9-03 relatability
0012  gate stays for the READER (identified only by the card's categories); the NARRATOR gets a full identity with
0013  her numbers. Unaware lane stays parked (Fish 8-31); write the PROBLEM-AWARE architecture.
0014  
0015  ## Alevia deltas ruled in (Fish 2026-09-03 evening, after the measured comparison
0016  `knowledge/formats/alevia-vs-dawn-v3-comparison-2026-09-03.md`)
0017  - **Clinician authority is allowed and wanted.** The authority character may be a nurse, a clinician, a
0018    specialist, with years stated, as in Alevia's cardiac nurse / cardiologist. Authorized dramatization; she
0019    never cites a study that is not ours, never names a drug, never says the product treats ADHD.
0020  - **Numbers in the results.** Counted, modest, qualified: "4 mornings out of 5", "11 school days without a
0021    yell", "the first time in 3 years". The week markers stay.
0022  - **Alevia paragraphs, not one sentence per line.** 120-180 paragraphs of 2-4 sentences, 15-35 words, one-line
0023    paragraphs only at punch moments (measured: all six longest-running Alevia copies, 24-28 words/paragraph).
0024    This overrides SOP §10's one-sentence-per-line for our lane.
0025  - **Narrator identity block with her numbers** (age, work, place, years of the problem) in the first
0026    paragraphs, and a **reader bridge** right after the hook ("I'm writing this because... if you have tried
0027    X, Y, Z... read this to the end"). Reader qualification is by what she TRIED, in her words, never by a
0028    product or drug list.
0029  - **Partial cast increase:** one more person who notices (her own mother or a friend), the kid speaks up to
0030    three times. Still no husband, no sibling (alienation rule).
0031  - **Quoted direct speech** with quotation marks; reported speech ("She said: ...") is the transcript register
0032    and is out.
0033  - **Durations on the failed stack** (how long each thing was tried); prices stay out of this lane.
0034  - **The close carries the Alevia rungs:** loss restate → the honest line ("I am not going to tell you it was a
0035    miracle. I am telling you what happened.") → guarantee line → the "If by..." conditional ladder → 👉 link →
0036    the "you can keep doing what you have been doing, or" fork → thank-you coda → final emotional line →
0037    first-name signature → P.S. (object) / P.P.S. (the honest count) / P.P.P.S. (small runs, do not wait) → 👉
0038    bare link last.
0039  - **Analogies:** one from the authority, one from the narrator repeating it, allowed.
0040  - The provenance-leak ban is narrowed to QA-speak ("I won't invent a review", "no fake before-and-after");
0041    the honest line is wanted.
0042  
0043  ## 1. Avatar depth (SOP §01) — the research file before the brief
0044  Answers, not guesses, from real people, in a single file the brief points at:
0045  - daily experience of the problem at the hour it happens · internal monologue (their exact phrases) ·
0046    failed attempts IN ORDER · their language (copied verbatim) · relationships (who notices, who picks up the
0047    slack, who said the thing that stung, who they fear disappointing) · the fear they will not say out loud.
0048  - Card slots added to every sub-avatar: BEFORE (who she was before it took over) · MONOLOGUE (3-5 verbatim
0049    self-talk lines) · WHO NOTICES · UNSPOKEN FEAR. Each slot is a quote with a source or it stays empty.
0050  
0051  ## 2. Awareness → architecture (SOP §02)
0052  - PROBLEM-AWARE (our lane): open on shared experience, the reader self-selects in the first 3 lines; failed
0053    attempts stack; the authority explains WHY nothing worked; alternatives eliminated; product = logical
0054    conclusion. 500-800 lines.
0055  - SOLUTION-AWARE: open on failed solutions, mechanism earlier, reveal sooner, more proof. 400-600 lines.
0056  - PRODUCT-AWARE: never long-format. Short, offer, proof.
0057  - A "line" is one sentence or fragment on its own line.
0058  
0059  ## 3. Mechanism (SOP §03)
0060  Root cause, not feature → the cascade (each link is a section) → why each partial solution fails (it addresses
0061  one link) → the product as the only thing that breaks the full cycle. Must be true, explainable over coffee in
0062  two minutes, a revelation to the reader, and lead to OUR product only. One analogy the authority speaks, that
0063  the reader will repeat to a friend. (Analogy is allowed HERE, spoken by the authority, and nowhere else.)
0064  
0065  ## 4. Story architecture (SOP §04) — line budgets for a ~700-line ad; scale proportionally
0066  | Part | Lines | Job | Must contain |
0067  |---|---|---|---|
0068  | 1 Hook | 1-3 | make them read line 2 | a moment, not a claim; no product/benefit/category; open loop; the reader recognizes her world. Problem-aware: she self-selects in 3 lines (our two hook families) |
0069  | 2 The World | 4-150 | make her feel SEEN | the BEFORE · the gradual decline as 3-5 small losses, each its own line · failed attempts named specifically and in order (prices where real) · the emotional toll (guilt, identity loss, quiet acceptance) · relationship impact (others notice, compensate, talk). No selling, no hinting. Short lines, white space, repetition for rhythm ("I tried X. Nothing.") |
0070  | 3 Inciting incident | 150-200 | make the status quo impossible | dramatic but believable, visceral, irreversible; introduces the AUTHORITY character |
0071  | 4 Mechanism | 200-400 | teach WHY she has it and WHY nothing worked | delivered through DIALOGUE; layered (root → cascade → why it compounds); narrator reactions between layers; one analogy; connect back to her life after each layer ("that is why the calm voice held until nothing moved"). Longest section after the world |
0072  | 5 Objection destruction | 400-500 | eliminate every alternative | narrator asks about each thing she tried, in the order she tried it; authority: acknowledge → the gap (it addresses part of the chain) → one memorable summary line. Never aggressive. The reader crosses them off herself |
0073  | 6 Product reveal | 500-550 (75-80%) | the natural resolution | the AUTHORITY recommends one specific named product; 2-3 specific differentiators that map to the mechanism; 2-3 paragraphs; stay in story voice; no marketing words |
0074  | 7 Future pacing | 550-650 | show her life after | WEEK-BY-WEEK: wk1 almost nothing · wk2 first small signal · wk3 something unexpected · wk4 measurable · wk6 other people notice · wk8 a modest, specific transformation that mirrors the inciting incident. Mix physical with emotional; end on the RELATIONSHIP RESOLUTION (the person from the hook says the thing) |
0075  | 8 Close | 650-700+ | convert investment into action | restate the stakes as LOSS · generational stake if real · risk reversal (the guarantee) · one direct CTA + link · final emotional line. Our offer floor rides here (brand + company intro + guarantee + link); the P.S. ladder is allowed after the close with a second link (Fish 8-31) |
0076  
0077  ## 5. Open loops (SOP §05)
0078  Never resolve all loops at once · open the next before closing the current · the main loop from line 1 spans at
0079  least half the ad · vary WHAT / WHY / HOW loops · loops at section transitions, not every line · loops open
0080  through events and spoken lines (our said-out-loud test still bans announcer narration).
0081  
0082  ## 6. Mechanism through dialogue (SOP §06)
0083  The authority asks permission to explain · physical beats break dense information · the narrator reacts ·
0084  the narrator asks the reader's questions · the authority speaks the analogy · PACING: 3-4 lines of mechanism
0085  to 1 line of reaction or physical beat. If it reads like a textbook, rewrite as conversation.
0086  
0087  ## 7. Objection handling disguised as story (SOP §07)
0088  Every alternative she tried or considered, most common first, most logical last. Acknowledge (it has merit) →
0089  gap (only part of the chain) → memorable summary. Tone sympathetic: "right instinct, wrong target."
0090  
0091  ## 8. Late reveal (SOP §08)
0092  Product name at 75-80% of the lines, never earlier. The authority recommends it. Differentiators are facts
0093  tied to the mechanism. Short. The transition is invisible: the last question is "so what do I actually do,"
0094  and the answer is the product.
0095  
0096  ## 9. Future pacing through results (SOP §09)
0097  Week-by-week, modest, specific, qualified; each week has a moment that makes her feel something; other
0098  characters notice; the relationship resolution is the emotional peak and comes last. Internal label:
0099  AUTHORIZED DRAMATIZATION when no buyer receipt exists (the SOP's own form); a verified Okendo/survey line
0100  may replace any week beat.
0101  
0102  ## 10. Formatting for Meta primary text (SOP §10)
0103  Sentences 5-15 words, rhythm varied · no emojis except 👉 before each link · no bold, no caps, no asterisks,
0104  no brackets. Block format for our lane is ruled above: Alevia paragraphs (2-4 sentences), one-liners at punch
0105  moments only.
0106  
0107  ## 11. The eight killers (SOP §11) — pre-publish check
0108  early reveal · shallow avatar (details that could apply to anyone) · rushed mechanism · tone switch at the
0109  reveal · overclaimed results · awareness mismatch · no stakes beyond the individual · weak inciting incident.
0110  
0111  ## 12. Process (SOP §12) and final check (SOP §13)
0112  Research → mechanism + outline → fast draft → four edit passes (cut, would-a-real-person-say-this, read
0113  aloud, find the ad switch) → read to someone who has the problem. Final check: line 1 loop · first 100 lines
0114  product-free · narrator sounds like the demographic · "this is MY story" specificity · mechanism in dialogue ·
0115  every alternative eliminated · invisible reveal · gradual believable results · loss-aversion CTA · would you
0116  keep reading.
0117  
0118  ## Our standing gates that ride on top (unchanged)
0119  Hook names her by symptom in lines 1-2 (two families) · relatability gate (no unverified reader qualifiers,
0120  no count on her repeats, narrator texture receipted) · no invented clinician, no invented science, no
0121  fabricated named testimonial with product outcomes · offer floor (brand at reveal, "Dawn Band is a small
0122  company and this is the only thing they make", "100% money-back guarantee. 60 nights.", "Trusted by 10,000+
0123  parents", link, CTA before and after the P.S. ladder) · never treats / cures ADHD · US spelling.

Native copy / references/long-native-clarity-and-jumble-gate.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/long-native-clarity-and-jumble-gate.md

SHA-256: b92a589929e75cf349d5468dfa000803bd0c909f1abaecbf3e11e935861fab7d

Full source · 79 lines
0001  # Long-Native Clarity and Final Jumble Gate
0002  
0003  Use after structural/SOP validation and before a native is called review-ready. Mechanical readability is necessary but not sufficient: a short sentence can still make no sense.
0004  
0005  ## 1. Failed-solution sequence
0006  
0007  Each attempt gets its own beat in this order:
0008  
0009  1. what she tried;
0010  2. why she thought it would help;
0011  3. what literally happened;
0012  4. why she stopped trusting it.
0013  
0014  Do not compress multiple methods into fragments such as `X came next`, `I laid it on`, `both helped`, or `neither stopped`. Do not open with `I tried everything the posts said`; name the concrete attempt.
0015  
0016  Passing shape:
0017  
0018  > I started getting up earlier because I thought I was the problem. I wanted to be ready and calm before he got up. I still ended up yelling by the door.
0019  
0020  ## 2. Paragraph-level jumble check
0021  
0022  Read every paragraph, not only the section recently edited. Flag:
0023  
0024  - chronology that jumps backward without a marker;
0025  - two unrelated actions in one beat;
0026  - unclear `it`, `that`, `this`, or `under it` references;
0027  - two speakers in one paragraph when the turn is hard to follow;
0028  - a method named without why she tried it;
0029  - a result stated before the action that caused it;
0030  - repeated week labels or duplicated transition lines;
0031  - count fragments whose subject must be reconstructed.
0032  
0033  Split speaker turns when necessary. One paragraph should perform one clear story job.
0034  
0035  ## 3. Semantic directness check
0036  
0037  Replace compressed copywriter abstractions with visible action or literal thought.
0038  
0039  | Reject | Prefer |
0040  |---|---|
0041  | `He was just not started.` | `He stood beside his shoes without reaching for them.` |
0042  | `He can hear you. He cannot begin.` | `He hears you, but his body still does not move.` |
0043  | `Your calm is his calm.` | `He hears the tension in your voice, then you get more tense when he does not move.` |
0044  | `The first thing he does is his own.` | `You do not have to give the first order from the doorway.` |
0045  | `He comes up with a clearer head.` | `He wakes with a clearer head.` |
0046  | `He gets lost between fewer steps.` | `He loses track of fewer steps.` |
0047  | `See what the rest of the hour does.` | `See whether you reach the door without yelling.` |
0048  
0049  If a reader must translate the sentence, it fails even if sentence length and grade score pass.
0050  
0051  ## 4. Final SOP read
0052  
0053  Run the mechanical gate, then answer the SOP final checks by hand:
0054  
0055  - line-one loop;
0056  - first third product-free;
0057  - narrator sounds like the demographic;
0058  - this-is-my-story specificity;
0059  - mechanism delivered through dialogue;
0060  - every narrator-specific alternative addressed;
0061  - reveal feels like story resolution;
0062  - results gradual and believable;
0063  - close uses loss aversion;
0064  - would the target reader continue?
0065  
0066  Also check the eight killers. A deterministic PASS does not override an editorial jumble finding.
0067  
0068  ## 5. Durable bans for this lane
0069  
0070  - no coffee filler;
0071  - no switching among `thread`, `forum`, and `post` in consumer copy—pick one channel;
0072  - no specific failed solution used as a universal reader qualifier;
0073  - no vague initiation shorthand;
0074  - average sentence under 14 words, no sentence over 28;
0075  - one exact sourced proof line may remain mechanically awkward only when the contract requires it; make the surrounding explanation explicit.
0076  
0077  ## Verification receipt
0078  
0079  Record final word/paragraph counts, reveal percentage, sentence average/max, grade estimate, clone check, receipt-table result, and a concise list of semantic repairs. Re-run the full gate after the final edit, not before it.

Native copy / references/master-brief.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/master-brief.md

SHA-256: fd399687215eb98086f327be5b6cf258a95fc9d32659a25e10b1ddc216fe31ac

Full source · 450 lines
0001  # NATIVE AD COPYWRITING MASTER BRIEF
0002  **Purpose:** HOW to write DawnBands native ad copy. Rules, format, structure, creative architecture. This file does NOT contain avatar knowledge — that lives in separate avatar docs. Read the avatar doc for your target audience BEFORE writing.
0003  **Date:** 2026-03-08
0004  
0005  ---
0006  
0007  ## FORMAT RULES (NON-NEGOTIABLE — CHECK EVERY LINE)
0008  
0009  - **1 sentence per line.** Default is always 1.
0010  - 2 sentences per line ONLY if the flow genuinely reads better that way.
0011  - **Short fragments: separate for emotional weight, clumped for action sequences.** "Eyes open. Turned it off. Walked back." = one fluid motion, clump on one line. But "The knocking. / The yelling. / The door slam." = each line is its own emotional punch, keep separate. If the fragments build rhythm or tension, they breathe on their own lines. If they describe a sequence of events, they flow together.
0012  - **NEVER 3+ sentences on one line.**
0013  - Period = line break. Always.
0014  - No em-dashes. Ever. Use periods + new lines.
0015  - No emoji on CTAs except 👉 for links.
0016  - Dialogue on its own line. Always isolated.
0017  - After writing all copy, do a FINAL PASS: scan every single line. If any line contains more than 2 sentences, split it.
0018  
0019  **Correct:**
0020  > I went in four times yesterday.
0021  > By the third time my voice was different.
0022  > By the fourth I was someone I don't want to be.
0023  
0024  **Wrong:**
0025  > I went in four times yesterday. By the third time my voice was different. By the fourth I was someone I don't want to be.
0026  
0027  ---
0028  
0029  ## META AD STRUCTURE (CRITICAL — understand this or the ad fails)
0030  
0031  **Primary text = the ad. She reads it FIRST when scrolling.**
0032  - The hook/WTF line MUST be the first line of primary text.
0033  - Never waste your best scroll-stopping line anywhere but line 1 of primary text.
0034  - **Lead with the most INTRIGUING element, not the chronological start.** Find the line that creates the most curiosity or emotional reaction and open with it — even if it's out of order in the story. "My husband thought I lost my mind" > "Six alarm clocks." because she hasn't explained WHY yet. The reader HAS to keep going.
0035  - **Never scene-set in line 1.** No "She stopped me in the driveway." No "I did the math once." No inventory lists. No context. The hook should CREATE a question, not answer one.
0036  - **Quotes hit hardest.** Direct dialogue as line 1 ("Do you need help? I hear screaming from your house every morning.") drops the reader into the middle of a moment with zero warmup.
0037  
0038  **Headline = bottom of the ad. Below the image. She reads it LAST.**
0039  - Use for: reinforcement, CTA summary, or a second punch.
0040  - NOT the hook. She already read the whole ad before she gets here.
0041  
0042  ## HEADLINE RULES (Meta `title` field)
0043  
0044  Every copy variant needs its own headline. This appears below the image in the ad.
0045  
0046  **Rules:**
0047  - Reinforces the body copy angle — NOT the hook (hook is in primary text)
0048  - First person voice preferred — sounds like her, not a brand
0049  - Sentence case. Not title case. Not ALL CAPS.
0050  - No emoji in headlines
0051  - Long headlines OK — Meta truncates but the full text shows on click
0052  - Must NOT duplicate any headline from another variant in the same set
0053  
0054  **Headline by variant type:**
0055  - **Story/Resignation:** The specific moment or realization. "I Did Everything His Therapist Told Me. He Still Couldn't Wake Up."
0056  - **Objection Killer:** The skepticism + the twist. "I've Spent $300 On Alarms. He Slept Through Every One. Then I Tried This."
0057  - **Pain-First → Mechanism:** The mystery/revelation. "His Medication Wears Off at 4pm. By 6:30am He's Unreachable. Nobody Explained That Cycle to Me."
0058  - **Consequence/Future Fear:** The stakes. "He Turns 18 in Two Years. Nobody Is Going to Stand Outside His Dorm Room."
0059  - **Risk Reversal + Social Proof:** The proof + offer. "9,000+ Reviews. 60-Night Guarantee. The Only Alarm Built For ADHD Brains."
0060  
0061  ---
0062  
0063  ## PRODUCT
0064  
0065  ### Offer-truth precedence
0066  Before writing, read the current validated brief and canonical product-truth file. Those sources override this reference and every historical example. Never inherit an old price, guarantee, app/setup claim, destination, or mechanism claim from a sample ad. When the current DawnBands offer is specified, preserve the token order `100% wake-up guarantee. 60 nights. $49.` while placing each sentence or fragment on its own line.
0067  
0068  - **Dawn Bands** — silent vibrating wrist alarm
0069  - $49 | 100% wake-up guarantee | 60 nights
0070  - Black silicone with ribbed diagonal lines; no visible screen or buttons; no app or subscription
0071  - Standalone band that places a vibration cue directly on the wrist. Describe a different sensory channel and signal placement, not a stronger signal.
0072  - Say `harder to sleep through`, never impossible or guaranteed every time.
0073  - Main PDP: dawnbands.com/products/wake-up-band-for-teens
0074  - ADHD PDP: dawnbands.com/products/wake-up-band-adhd
0075  - **Main listicle:** dawnbands.com/pages/7-reasons-teen-sleeps-through-alarms
0076  - **ADHD listicle:** dawnbands.com/pages/adhd-teen-alarm
0077  
0078  ---
0079  
0080  ## THE MECHANISM (load the canonical selling doctrine)
0081  
0082  Before writing DawnBands copy, read `/Users/ayden/.openclaw/workspace/knowledge/brands/dawnbands/selling-root-cause-mechanism.md`. It governs this section.
0083  
0084  The default selling chain is **Deep-Sleep Signal Lock → Somatic Wake Bypass™ → Wrist-First Handoff™**:
0085  
0086  1. adolescent circadian delay keeps biological night running later
0087  2. accumulated sleep debt increases the pressure holding him asleep
0088  3. the sleeping brain gates/suppresses repeated room sound
0089  4. sleep inertia lets his body silence alarms or answer before conscious initiation and memory return
0090  5. Mom becomes the emergency pathway that finally breaks through
0091  6. DawnBands bypasses the muted sound pathway with a tactile signal on the wrist
0092  7. the first cue begins on him before it ends in her voice
0093  
0094  Use aggressive biological simplification to sell the mechanism. Do not flatten the copy to “sound and vibration are different.” Preserve separate substantiation discipline for exact-SKU superiority, fabricated testimonials, customer timelines, offer facts, and attributed proof.
0095  
0096  **The mechanism belongs in the MIDDLE of long-form problem-aware copy. NEVER use science or product mechanics as the opening hook.** Compressed solution-aware/product-aware formats may reveal the mechanism earlier when the validated brief explicitly calls for it.
0097  
0098  **Attach the canonical villain:** the repeated room-sound rescue system. It keeps turning up the pathway his sleeping brain already suppresses, then outsources the final wake-up job to Mom. Redirect blame away from both the teen and parent.
0099  
0100  ---
0101  
0102  ## AVATAR DOCS (READ BEFORE WRITING — MANDATORY)
0103  
0104  **The avatar docs contain the raw quotes, exact language, interior world, and copy triggers for each target audience. This brief tells you HOW to write. The avatar docs tell you WHO you're writing to and WHAT words to use.**
0105  
0106  **⚠️ DEFAULT TO THEIR WORDS OVER YOUR OWN — always.** When you're reaching for a word, check the avatar doc's language bank FIRST. If their word exists for what you're trying to say, use theirs. Your instinct will be to write something "better" — don't. Their words convert. Yours don't.
0107  
0108  **Read the avatar doc for your assigned target BEFORE writing a single line.**
0109  
0110  
0111  > ⭐ **SINGLE AVATAR RULING (Fish 2026-09-02).** Every DawnBands brief targets the mother of a 13–18 ADHD teen. Load `knowledge/brands/dawnbands/adhd-mom-lockin.md` FIRST, then `dawnbands-avatar-adhd-teen-mom.md`. The "Teen Mom (primary)" generic row below is PARKED — do not brief it. ADHD is named in the hook. Survey basis is n=138 (`survey-intel-post-purchase-2026-09-01.md`).
0112  
0113  | Avatar | Doc Path | Who They Are |
0114  |---|---|---|
0115  | **Teen Mom** (primary) | `knowledge/dawnbands-avatar-interior-2026.md` | Mom 38-49, teen 14-17 who can't wake up. She's become his alarm clock. |
0116  | **ADHD Mom** | `knowledge/dawnbands-avatar-interior-2026.md` (Part 7) | Same mom, but teen has ADHD. More informed, more exhausted, more isolated. |
0117  | **ADHD Adult Self-Buyer** | `knowledge/dawnbands-avatar-adhd-adult.md` | Adult 25-45 with ADHD buying for themselves. Shame, career consequences, exec function wall. |
0118  | **Partner/Spouse Buyer** | `knowledge/dawnbands-avatar-partner-spouse.md` | Partner buying because THEIR sleep is ruined by the heavy sleeper's alarms. |
0119  | **Broad Deep Sleeper** | `knowledge/dawnbands-avatar-broad-deepsleeper.md` | Adult heavy sleeper (any age) buying for themselves. Shift workers, college students, professionals. |
0120  
0121  **Victory = ABSENCE, not presence.** Across ALL avatars. She/he doesn't celebrate. They notice what STOPPED happening. Write victory as what disappeared, not what arrived. "I didn't yell today." "He hasn't missed a shift in six weeks." "I slept through the night for the first time in three years."
0122  
0123  ---
0124  
0125  ## ZAK FRAMEWORK — NATIVE AD STRUCTURE
0126  
0127  **Line-by-line structure:**
0128  1. **Hook (first 3-5 lines):** Pattern interrupt. NOT an ad. Use her exact language. Must pass Gate 1-3 (interrupt → earn attention → recognition).
0129  2. **Story opening:** First person, raw. Start in a specific moment. Small detail, not summary. ("I pulled his covers at 6:42" not "mornings were hard")
0130  3. **Agitation:** Surface the daily emotional weight. "Tried Everything" stack. Breaking point moment.
0131  4. **Root cause / mechanism:** NOT the product. The villain (sound). The explanation. One analogy max. Short punchy sentences. **This goes in the MIDDLE.**
0132  5. **Product reveal + CTA:** "Of course. This is what I've been looking for." CTA goes HERE — immediately at product reveal.
0133  6. **Close:** Concrete outcome. Risk reversal. Guarantee.
0134  
0135  **Hook formulas (match to awareness level):**
0136  - Formula 1: PAIN → VALIDATION → INTRIGUE
0137  - Formula 2: STORY → TENSION → CLIFFHANGER  
0138  - Formula 3: QUESTION → TWIST → PROMISE
0139  - Cold Meta traffic = Levels 1-2 awareness. Meet her there.
0140  
0141  **Cadence rule:** Alternate Long (build) → Short (punch) → Short (promise) → Long (explain) → Short (reveal). Read out loud. If you stumble, rewrite.
0142  
0143  ---
0144  
0145  ## TARGETING — WHICH LISTICLE + PDP
0146  
0147  **4 listicles, 1 PDP. Route based on avatar + angle:**
0148  
0149  | Avatar/Angle | Listicle | PDP |
0150  |---|---|---|
0151  | Teen mom (general) | `dawnbands.com/pages/7-reasons-teen-sleeps-through-alarms` | `/products/wake-up-band-adhd` |
0152  | ADHD mom (parent voice) | `dawnbands.com/pages/adhd-teen-alarm` | `/products/wake-up-band-adhd` |
0153  | ADHD adult (self-buyer) | `dawnbands.com/pages/adhd-adult-alarm` | `/products/wake-up-band-adhd` |
0154  | Broad deep sleeper / partner / spouse | `dawnbands.com/pages/why-you-sleep-through-alarms` | `/products/wake-up-band-adhd` |
0155  
0156  **Every new ad gets duped to BOTH listicle + PDP.** Same creative, two destinations. Meta picks the winner.
0157  
0158  **Routing rules:**
0159  - Any ad mentioning ADHD, medication, executive function → ADHD listicle (mom or adult version based on voice)
0160  - Any ad about teens who can't wake up (no ADHD) → Teen mom listicle
0161  - Any ad from adult's own perspective (shift worker, college, professional, partner) → Broad listicle
0162  - When in doubt → Broad listicle (it's the most general)
0163  
0164  ---
0165  
0166  ## 4 VARIANT TYPES (write one of each per ad)
0167  
0168  **⚠️ OBJECTION KILLER = TOP PERFORMER (Mar 2026 data)**
0169  Live data from Teen Mom Natives campaign shows Objection Killer variants converting across multiple ads:
0170  - teen-mattress-retool V2 (objection): $9.55 CPP ← best in campaign
0171  - teen-relationship-restoration V2 (objection): $6.53 CPP
0172  - teen-identity-erosion V2 (objection): $49.93 CPP but only variant converting
0173  - teen-teens-voice V1 (objection-style): $19.62 CPP
0174  **Implication:** She's skepticism-first, story-second. Even story variants should open with a beat of "I know, I know, another thing" energy before going into the narrative. The drawer of failed products is her entry point.
0175  
0176  **1. Story/Resignation Arc**
0177  - First person narrative. The mom who tried everything.
0178  - Entry point: a specific failed moment or the resignation breaking point.
0179  - **Consider opening with a skepticism beat even in story format** — she resonates with "I wasn't looking for this" energy.
0180  - Emotion: exhaustion → discovery → quiet relief.
0181  - Awareness: Level 2 (Problem Aware).
0182  
0183  **2. Objection Killer** ← PRIORITY VARIANT, write 2 per ad if possible
0184  - Leads with the skepticism she already has.
0185  - "Won't he just sleep through this too?" / "Another alarm?" / "$49 for a vibrating wristband?"
0186  - Disarms before selling. Names the drawer of failed products.
0187  - Emotion: cynicism → surprise → belief.
0188  - Awareness: Level 3 (Solution Aware).
0189  
0190  **3. Pain-First → Mechanism Reveal**
0191  - Opens with vivid daily pain (time-stamped morning sequence, sensory detail).
0192  - Mechanism lands in the middle as the "nobody explained this" revelation.
0193  - Emotion: confusion/frustration → clarity → hope.
0194  - Awareness: Level 2 (Problem Aware).
0195  - **DO NOT open with mechanism. Open with pain.**
0196  
0197  **4. Consequence/Future Fear**
0198  - What happens if nothing changes. College, jobs, independence, relationship.
0199  - The clock is ticking. Urgency from reality, not marketing.
0200  - Emotion: fear → urgency → action.
0201  - Awareness: Level 2-3.
0202  
0203  ---
0204  
0205  ## VARIANT DIVERSITY RULES (MANDATORY — prevents Andromeda overlap)
0206  
0207  ### Mechanism Delivery Must Vary Across Variants
0208  The mechanism (auditory channel closes → touch stays open) appears in every variant. **It must be delivered differently each time.**
0209  
0210  **Do NOT copy-paste the same mechanism paragraph across 5 variants.** A real person wouldn't explain it the same way twice.
0211  
0212  Mechanism delivery styles (use a DIFFERENT one per variant):
0213  1. **Full science** — "The sleeping brain suppresses the auditory channel..." (max 1 variant per ad)
0214  2. **Casual/vague** — "I read something about how sound doesn't actually reach deep sleepers. Something about the brain closing a door." (no jargon)
0215  3. **Someone else explained it** — "My sister sent me an article. Apparently the brain just... blocks sound when you're that deep." (secondhand)
0216  4. **Barely explained** — "It vibrates instead of making noise. I don't fully understand why that's different but it is." (outcome only)
0217  5. **Analogy** — "It's like knocking on a soundproof door vs. tapping someone on the shoulder." (metaphor)
0218  6. **Skip entirely** — Some variants (especially social proof V5) can just say "it works differently than sound alarms" and move on.
0219  
0220  ### CTA Format Must Vary
0221  Do NOT use "👉 Here's [reason/explanation]:" for every variant. Real people link differently.
0222  
0223  CTA styles (rotate):
0224  - "👉 Here's what I read:" + URL (max 2 per ad)
0225  - "I'll put the link below." (no 👉)
0226  - "Link is here if you want to read it:" + URL
0227  - "The explanation is at [URL]" (embedded naturally)
0228  - "Someone posted this in my group:" + URL
0229  - No mid-copy CTA at all — just the close CTA
0230  
0231  **CTA must flow with the story, not feel templated.** Do NOT copy-paste "$49. 60 nights. Full refund." on every variant. Weave guarantee/price into the narrative naturally:
0232  - "I had 60 nights to send it back. Never even considered it."
0233  - "$49. Full refund if it joins the drawer. It didn't."
0234  - Some variants skip PRICE entirely — the number is optional (Fish 2026-08-31).
0235  - Some variants bury the guarantee in a callback line, not a bullet list.
0236  If the CTA block reads like it was pasted onto 5 different stories, rewrite it.
0237  
0238  ⚠️ **OFFER FLOOR (Fish 2026-08-31 — overrides the variation permissions above).** Varying the CTA never means DELETING the close. Every body, every length, every destination carries: brand named at the reveal, the guarantee stated plainly, a 2-4 line company/credibility intro, and at least one link. Price is the only optional element. A body with no brand, no offer, and no link is a click generator, not a native — receipt: v11-a-unassigned-hour, 4.13% link CTR at $0.92 CPLC, $46-56 CPA. Full rule + close-block shape: SKILL.md §3a.
0239  
0240  ### Social Proof Variants (V5) Must Sound Real
0241  - Reviews should have typos, incomplete thoughts, weird phrasing
0242  - Not every review is 5 stars. Mix in a 4-star that still converts.
0243  - Names should feel real (not "Trish M." and "Kara D." — too clean)
0244  - Include at least one review that sounds like it was written at 2am on a phone
0245  - Reference platform ("posted this in an ADHD group", "saw this on Amazon") to ground it
0246  
0247  ### Body Copy Length Variation (CRITICAL — Mar 2026 data)
0248  **Both long AND short are converting.** n6-medication original (706 words) shows signal. Short teen mom copies also cooking hard. DO NOT default to one length.
0249  
0250  **Per ad, aim for this mix:**
0251  - 2-3 variants LONG: 500-700 words
0252  - 2 variants SHORT-MEDIUM: 150-300 words (punchy, quick, scroll-stop-and-convert)
0253  - The Objection Killer variants can be either — some of the best performers are medium-length objection openers
0254  - Let Meta test which length wins per angle
0255  
0256  **⚠️ "LONG" does NOT mean dense paragraphs or walls of text.** Long variants still use 1-2 sentences per line with lots of line breaks. The length comes from MORE lines, not longer paragraphs. Think: a long Facebook post that's easy to scroll through, not an essay. Even ADHD moms need to be able to scan it. The 1-sentence-per-line format rule applies to all **Mode A** variants regardless of total word count — "LONG" in this section means 500–700-word Mode A copy. **It does NOT apply to Mode B long-form witness natives**, which run PARAGRAPH prose per the 2026-07-31 ruling (`dawnbands-long-form-witness-native.md` → Formatting; measured basis in `alevia-live-anatomy-2026-08-12.md`). **Long-format natives are governed by SOP 2.0 as of 2026-09-03** (`long-format-native-sop-v2-contract.md`: eight parts, Alevia paragraphs of 2-4 sentences, clinician authority in dialogue, counted week-by-week results, reveal 65-82%, close rungs + P.S. ladder); that contract wins over this section wherever they disagree.
0257  
0258  **Do NOT write all 5 variants at the same length.** Variety in length = more data on what converts per angle.
0259  
0260  ---
0261  
0262  ## STRUCTURAL CADENCE PATTERNS (from proven winners — Uflora, True Nutra, Curvani)
0263  
0264  These patterns are extracted from gold-standard native ads that convert at 9 figures. Subagents MUST use these rhythms. Don't invent your own cadence — follow these.
0265  
0266  ### Repetition / Anaphora
0267  Build rhythm through repeated structure on separate lines:
0268  ```
0269  A smell that followed me out of the bathroom.
0270  A smell I could catch through my jeans if I sat wrong.
0271  A smell that made me terrified to be in a car with the windows up.
0272  ```
0273  
0274  ### Internal Monologue Stacking
0275  Each thought on its own line. Short. Obsessive. Circular:
0276  ```
0277  Can she smell me right now?
0278  Can she smell me?
0279  Oh God, can she smell me?
0280  ```
0281  
0282  ### Failed Solutions List
0283  Short. Stacked. Each its own line. Include price and verdict:
0284  ```
0285  Summer's Eve wipes. Worse.
0286  Boric acid suppositories. Burned like hell. Came right back.
0287  Antibiotics from my doctor. Ten days of relief. Then worse than before.
0288  ```
0289  
0290  ### Dialogue Isolation
0291  Dialogue ALWAYS gets its own line. Never buried in a paragraph:
0292  ```
0293  She looked at me and said:
0294  
0295  "You're not unhygienic.
0296  Your microbiome has collapsed."
0297  ```
0298  
0299  ### Trigger Scenarios
0300  Question + answer, each its own line. Shows the pattern:
0301  ```
0302  After sex?
0303  Semen temporarily raises vaginal pH. Gardnerella feeds.
0304  
0305  During your period?
0306  Blood contains iron. Gardnerella's favorite fuel.
0307  ```
0308  For DawnBands: "Field trip? She overslept. Test day? She overslept. First day of school? She overslept."
0309  
0310  ### Timestamped Transformation
0311  Specific timestamps for the victory section. NOT "after a few weeks." Concrete:
0312  ```
0313  Day 7: I got through an entire morning without checking once.
0314  Not once.
0315  
0316  Day 9: Had sex. Lay there afterward, waiting.
0317  Like you wait for thunder after lightning.
0318  
0319  Nothing.
0320  ```
0321  For DawnBands: "Day 3: He woke up before I did. I lay there listening to footsteps in the hallway."
0322  
0323  ### The External Witness
0324  Someone else notices the change BEFORE the protagonist announces it. Short. Devastating:
0325  ```
0326  My husband noticed before I told him.
0327  
0328  "You're lighter," he said.
0329  
0330  Not my weight.
0331  ```
0332  For DawnBands: "His teacher emailed me. 'Whatever you changed, keep doing it. He's been on time every day this week.'"
0333  
0334  ### The "Waiting for It to Fail" Beat
0335  After the product works, she waits for it to stop working. This validates her skepticism and makes the victory real:
0336  ```
0337  Like you wait for thunder after lightning.
0338  
0339  Nothing.
0340  ```
0341  For DawnBands: "I waited for the text from school. It didn't come. Not that day. Not the next."
0342  
0343  ### The "No X. No Y. No Z." Denial Rhythm
0344  Stack what DIDN'T happen. Powerful for transformation sections:
0345  ```
0346  No belly bulge. No puff. No 'why am I gaining everywhere but my booty?'
0347  ```
0348  For DawnBands: "No yelling. No third trip upstairs. No 'you're going to be LATE.'"
0349  
0350  ### The P.S. Emotional Bomb
0351  The most emotionally devastating line goes in a P.S. at the end. It hits harder because it feels like an afterthought — something too vulnerable for the main copy:
0352  ```
0353  💡 P.S. He told me once — in the car, not looking at me — that he hates himself every morning that I have to come in. He was thirteen. I still think about that.
0354  ```
0355  
0356  ---
0357  
0358  ## CREATIVE ARCHITECTURE (from 8-9 figure operators)
0359  
0360  ### PDA Framework (Persona × Desire × Awareness)
0361  When writing 5 variants, don't just swap headlines on the same angle. Each variant should vary across at least ONE of these dimensions:
0362  - **Persona**: Who is telling the story? (exhausted mom, guilty dad, the teen herself, a teacher, a neighbor)
0363  - **Desire**: What does she want? (calm mornings, relationship repair, independence for him, her own identity back)
0364  - **Awareness**: Where is she? (doesn't know vibrating alarms exist, knows but skeptical, tried competitors and failed)
0365  If all 5 variants target the same persona with the same desire at the same awareness level, you've written 1 ad 5 times.
0366  
0367  ### Emotional Voltage
0368  An angle is NOT an emotion. An angle is a belief system/story that contains 3-5 emotional triggers in sequence. The triggers build cumulative voltage — each line adds charge until the reader HAS to act.
0369  - **Low voltage**: "Mornings are hard." (one flat emotion)
0370  - **High voltage**: Shame (I'm the yelling mom) → Fear (he'll never be independent) → Guilt (he hates himself for it) → Hope (but what if) → Relief (he walked downstairs on his own)
0371  Pick ONE primary emotion to build to peak. Layer supporting triggers to increase voltage. Don't scatter — BUILD.
0372  
0373  ### "Make Inaction More Dangerous Than Action"
0374  The reader must feel that NOT clicking is riskier than clicking. The copy should make doing nothing feel like a choice with consequences:
0375  - "In three years he starts college. Who's going to wake him up?"
0376  - "847 mornings. That number only goes up."
0377  - "Every morning I don't fix this, I'm teaching him he can't do it without me."
0378  
0379  ### Identity Loss = Meta-Pattern, Not an Emotion
0380  Identity loss ("I'm not the mom I wanted to be") isn't a single emotion to trigger — it's the REASON multiple emotions appear simultaneously. Don't write "she feels identity loss." Write the specific emotions that identity loss causes: shame, grief, anger at herself, longing for who she used to be. Channel the pattern through specific moments, not labels.
0381  
0382  ### Emotional State First, Then Demographics
0383  Don't start by writing to "mom 38-55." Start by finding the emotional state that converts (resignation, shame, exhaustion, fear of permanent damage). THEN layer in demographic-specific scenarios. The emotional state is shared across similar demographics — that's why it scales. The scenario makes it feel personal.
0384  
0385  ### Mechanism = Highest Leverage Variable
0386  Same product, same funnel, same traffic — changing the mechanism explanation alone has driven operators from breakeven to $800k/month. The mechanism is not "feature X." It's the WHY behind why everything else failed and why THIS works.
0387  For DawnBands: "Sound goes through air. His brain filters air. Touch goes through nerve endings. His brain can't filter nerve endings." — THIS is the mechanism. If it's missing or weak, nothing else saves the ad.
0388  
0389  ### Mechanism Reveal Structure (4 beats)
0390  1. **Setup (create the gap)**: Acknowledge her pain, list what she's tried, highlight why it all failed. Don't reveal why yet.
0391  2. **Reveal (the insight)**: The hidden truth — teen brains suppress SOUND during deep sleep. It's biological, not behavioral.
0392  3. **Connection (link to HER)**: This is why the Sonic Boom failed. This is why 5 alarms failed. This is why SHE had to become the alarm.
0393  4. **Solution bridge (inevitable product)**: Touch bypasses the auditory channel entirely. A vibration on the wrist can't be filtered. DawnBands.
0394  
0395  ---
0396  
0397  ## ⛔ NEVER DO (CHECK EVERY VARIANT AGAINST THIS LIST BEFORE SUBMITTING)
0398  
0399  This is the most important section. Violating ANY of these = rewrite the variant from scratch.
0400  
0401  ### HOOKS — NEVER
0402  - ❌ NEVER open with mechanism / science / "the brain does X" — DEAD ($20 spend, 0 purchases)
0403  - ❌ NEVER scene-set in line 1. No "She stopped me in the driveway." No "I did the math once." No "I'm sitting in the drop-off line." Hook = INTRIGUE, not context.
0404  - ❌ NEVER start chronologically. The most intriguing moment goes first, even if it happened in the middle of the story.
0405  - ❌ NEVER put your best line in the headline field. Headline = bottom of the ad. She reads it LAST. Your scroll-stopper goes in primary text line 1.
0406  - ❌ NEVER open with an inventory list. "Six alarm clocks. Two phone docks." is not a hook. "My husband thought I lost my mind." is.
0407  - ❌ NEVER use generic feature hooks ("silent alarm," "for heavy sleepers," "comfortable") — commodity, every competitor uses these.
0408  - ❌ NEVER use curiosity-click bait — native1 had 24.8% CTR, 0 purchases. Intrigue must lead to the sale, not just the click.
0409  
0410  ### LANGUAGE — NEVER
0411  - ❌ NEVER use: "game changer," "life changing," "revolutionary," "amazing," "incredible," "journey," "struggle," "battle," "warrior," "graveyard"
0412  - ❌ NEVER use weak modifiers: "very," "really," "just," "truly," "literally"
0413  - ❌ NEVER use em-dashes. Ever. Use periods + new lines.
0414  - ❌ NEVER use a word that feels like every AI would pick it. If it sounds "writerly" or poetic where a real mom would say something uglier, kill it. "The drawer I can't open without getting mad" > "the graveyard of failed alarms."
0415  - ❌ NEVER use polished, corporate-sounding testimonials. Reddit/Amazon voice converts. "Literally changed our lives ⭐⭐⭐⭐⭐" = death.
0416  - ❌ NEVER lead with reviews/testimonials. Defeats native format — looks like an ad instantly. Reviews = body, never the hook.
0417  
0418  ### FORMAT — NEVER
0419  - ❌ NEVER put 3+ sentences on one line. Max is 2 (and only when it flows better).
0420  - ❌ NEVER write dense paragraphs. "Long" = more lines with breathing room, NOT bigger paragraphs.
0421  - ❌ NEVER bury dialogue inside a paragraph. Dialogue gets its own line. Always.
0422  - ❌ NEVER write all 5 variants at the same word count. Mix long (500-700) and short (150-300).
0423  
0424  ### STRUCTURE — NEVER
0425  - ❌ NEVER put mechanism in the opening. Mechanism goes in the MIDDLE after the emotional setup.
0426  - ❌ NEVER copy-paste the same mechanism paragraph across variants. Each variant delivers the mechanism differently (full science, casual, secondhand, barely explained, analogy, or skip).
0427  - ❌ NEVER copy-paste the same CTA block across variants. Each CTA flows with its own story. No templated "$49. 60 nights. Full refund." on every variant.
0428  - ❌ NEVER reuse an analogy across variants. "Locked door" in V1 = can't appear in V2-V5.
0429  - ❌ NEVER write 5 variants with the same persona, desire, and awareness level. That's 1 ad written 5 times. Vary at least one dimension per variant.
0430  - ❌ NEVER write victory as what STARTED happening. Victory = ABSENCE. What STOPPED. "I didn't yell today." Not "Mornings are peaceful now."
0431  - ❌ NEVER skip the villain. If the copy doesn't blame sound alarms / the alarm industry / bad advice, she'll blame herself and freeze. No sale.
0432  
0433  ### AVATAR — NEVER
0434  - ❌ NEVER use generic avatar language ("stressed professional," "busy mom"). Use HER exact words from the VOC bank.
0435  - ❌ NEVER write to the wrong awareness level. Cold Meta traffic = Level 1-2. Don't write Level 4 copy to people who don't know the product exists.
0436  - ❌ NEVER blame or shame the avatar. She already feels terrible. The copy should make her feel understood, then redirect blame to the real problem (sound alarms).
0437  - ❌ NEVER say "sometimes" or "usually" — she says "every single morning." Always absolute.
0438  - ❌ NEVER future-pace vaguely. "You'll feel better" = nothing. "He walked downstairs on his own Tuesday" = conversion.
0439  - ❌ NEVER try to trigger all emotions at once. Pick ONE primary, build voltage through supporting triggers in sequence.
0440  
0441  ### DEAD ANGLES (CONFIRMED BY DATA — $0 REVENUE)
0442  - ❌ Mechanism as hook — $20 spend, 0 purchases
0443  - ❌ Physical letter/document as hook image — $38 spend, 0 purchases
0444  - ❌ Curiosity-click bait — 24.8% CTR, 0 purchases across 2 geos
0445  
0446  ---
0447  
0448  ## GOLD STANDARD EXAMPLE
0449  
0450  See the attached file `gold-standard-example.md` — the native8 teen mom test batch (Mar 7 2026). This follows ALL current rules including hook placement, CTA flow, fragment pairing, cadence patterns, and the ⛔ NEVER DO list. Use this as the quality bar.

Native copy / references/matrix-angle-execution-map.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/matrix-angle-execution-map.md

SHA-256: 85a73e7ad81491f98be0e5851fe4d86812a829073d028fa0d61105c9f63bc6bf

Full source · 93 lines
0001  # Matrix → execution map (2026-08-13)
0002  
0003  When Fish consults the ad-batch-hub testing matrix ("expand my tests", "feed cell X", "the
0004  grey cells") this file resolves each matrix ANGLE column into the writing execution: what the
0005  angle IS, its device basis, the DawnBands form, default mode/length, live receipts, and the
0006  claim constraints. Matrix concepts (static / native-short / native-long) select Mode A vs
0007  witness architectures as usual; this map governs the angle layer.
0008  
0009  Angle definitions are HOOK-level (the matrix scores angle off the hook — the ad's strategic
0010  bet), matching `build_matrix.py`. An ad's body may touch several angles; the hook decides.
0011  
0012  ## The original eight (established — one-liners)
0013  
0014  | Angle | The hook's bet | Default execution |
0015  |---|---|---|
0016  | mechanism | why sound fails / how waking actually works | Mode A mid-form; mechanism is Stage 4-5 saturated — never the differentiator, always the mid-copy efficacy answer |
0017  | reframe | "not laziness/discipline — biology" as a CLAIM | statics (winner skeleton) + Mode A; our #2 angle |
0018  | relatability | recognition punch — her morning, mirrored | any length; lines 3-8 VOC-specificity test decides life or death |
0019  | stakes | future-fear with a countdown (college, job, 18) | Mode A mid or compressed witness; milestone-rupture opener form folds here |
0020  | liberation | identity exit — "never be their alarm clock" | short/mid; our #1 angle |
0021  | proof | it worked — receipts, reviews, retro-proof | compressed witness (fable-v10 cell) or testimonial statics; see devices below |
0022  | usp | product-forward features/guarantee/objection-killer | Mode A short; objection-killer is the top Mode A performer |
0023  | offer | promo/price/occasion | short; brand-page checklist evergreen (T4) lives here |
0024  
0025  ## The three Alevia columns (added 8-12 — census-confirmed the complete transferable set)
0026  
0027  ### wrong-label — "Wrong-label reveal"
0028  **Is:** a narrated CHRONOLOGY of the label being wrong — years of treating it as character
0029  (lazy / defiant / unmotivated), then the verdict falls. Distinct from reframe: reframe states
0030  the claim; wrong-label lives the mislabel and confesses it.
0031  **Alevia basis:** wrong-diagnosis reveal, 50 copies/123 ads ("turns out I never had X at all").
0032  **Dawn form:** mislabel is BEHAVIORAL, never medical. The mom (or adult self) mis-graded the
0033  kid/themselves; the evaluation or realization re-reads history. Mislabel artifacts: charts,
0034  contracts, punishments enforced. The apology beat is this angle's strongest weapon.
0035  **Mode/length:** compressed witness or true Mode B; the chronology needs room — never a static.
0036  **Receipts:** u3-wronglabel-star-chart (+ modeb), u4 sleep-study (staged 8-12); newdoc-v1 ($38).
0037  **Gates:** the wrongness is about effort/character judgments, never a doctor's medical error;
0038  sleep-study-normal is grief, not malpractice.
0039  
0040  ### defiance — "Authority defiance"
0041  **Is:** villain = the dismissal chain; narrator stops WAITING for the system to own the problem.
0042  **Alevia basis:** treatment refusal, 44/93 ("tore up the prescription") — their form is medical
0043  refusal; OURS IS NOT.
0044  **Dawn form:** PROCEDURAL defiance only — "he'll grow out of it," natural-consequences advice,
0045  waitlists, the care plan with no morning section. Mom keeps the meds, keeps the appointments,
0046  defies the process by solving 6:30 herself without a referral.
0047  **Mode/length:** compressed witness / Mode B; the dismissal chain is a chronology.
0048  **Receipts:** u5-defiance-waitlist (+ modeb, staged 8-12).
0049  **Gates:** never anti-medication, never anti-doctor as people; zero treatment/efficacy claims;
0050  the defiance object is a process (hold music, handouts, blank page four), not a treatment.
0051  
0052  ### witness — "Witness POV"
0053  **Is:** the narrator is a third party OUTSIDE the buyer-sleeper dyad — dad watching wife+son,
0054  teacher watching cohorts, observed-peer ("a mom I know"). NOT the managing mom narrating her
0055  own kid: that is the witness FORMAT (Mode B's default stance) and classifies by its hook angle.
0056  **Alevia basis:** spouse-witness at volume (their most-used device, 40/90 third-party-coded).
0057  **Dawn form:** two generations of it now — (1) single-scene observer (old: dad-614am recording,
0058  r2-d2 dorm-call peer), (2) witness-as-buyer: the observer runs the shift, fails, researches,
0059  BUYS — full DR arc carried by the witness (new: u1 dad-legal-pad). Teacher = longitudinal
0060  occupational witness, authority via observation, product enters through the mom's dialogue.
0061  **Mode/length:** compressed witness or Mode B.
0062  **Receipts:** dad-614am $19.63 @ 6.67 (thin, positive), r2-d2 $514 @ ~1.55, km08 $37;
0063  u1/u2 (+ u1-modeb) staged 8-12.
0064  **Gates:** the witness never fabricates medical authority; occupational witnesses speak from
0065  observation, and any diagnosis mention is carried with the family's stated permission in-story.
0066  
0067  ## Proof-column DEVICES (census 8-13 — these are not new columns)
0068  
0069  - **Comparative two-person proof** (Alevia 25/63): side-by-side lives — "her son wakes himself
0070    to a wrist tap / mine gets a mother." Two kids, two houses, same alarms, one variable.
0071    UNTESTED for Dawn; classifier now files it to proof. Pilot-grade device.
0072  - **Aggregated survey voice** ("we asked 40 buyers…"): white space even vs Alevia (they run it
0073    only as a PDP block, zero ads). Files to proof. Real survey data only (n=40 post-purchase),
0074    never invented percentages.
0075  - **Retro-proof** (fable-v10 / S2 receipts): victory-as-absence receipts (zero missed 8ams,
0076    empty email folder, lobby-call text). The proven core of the column.
0077  
0078  ## Devices that are NOT matrix angles (skill-side, any column)
0079  
0080  Guilt-confession opener · discovery-chain reveal · P.S. ladder close · rupture-object cold
0081  open · milestone rupture (files to stakes). See `alevia-live-anatomy-2026-08-12.md`.
0082  
0083  **Deliberately absent from the matrix:** death-stakes (Alevia's biggest payload, 175/331 —
0084  banned register for this brand); medical bylines/clinic authority; teen self-voice.
0085  
0086  ## Consult flow (when Fish says "use the matrix to expand tests")
0087  
0088  1. Pull the live matrix (hub #matrix or `site/matrix/data/matrix.json`) — grey/thin cells in
0089     ACTIVE lanes (parent-adhd, parent-teen, adult-self) are the test surface; pixel ruling says
0090     on-avatar first.
0091  2. Resolve cell → this map for angle execution + mode; concept axis picks Mode A vs witness.
0092  3. Row-per-ad diversity matrix, pilot, surgical pass, per the SKILL workflow. Every artifact/
0093     dollar/analogy/victory belongs to exactly one ad in a batch.

Native copy / references/mode-b-first-draft-checklist.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/mode-b-first-draft-checklist.md

SHA-256: 5f3c292c419657613243f8397bf8bb2a77dcc6334326ba98c1752b581c2dd4b8

Full source · 64 lines
0001  # Mode B first-draft checklist (DawnBands)
0002  
0003  Use with `dawnbands-long-form-witness-native.md`. This is the short operator card for drafting, not a second doctrine.
0004  
0005  ## Before prose
0006  
0007  - [ ] Brief intentionally selected Mode B (not assumed from “write a native”)
0008  - [ ] Narrator mode declared — default **ownership-voice** if no verified buyer receipt
0009  - [ ] Motif chosen (one only)
0010  - [ ] Primary open loop + ≥1 secondary loop
0011  - [ ] Inciting pressure = RECEIPT or HYPOTHETICAL_OWNERSHIP (never dramatized fake past)
0012  - [ ] Exact-SKU claims split: VERIFIED / SELLER-CLAIMED / UNKNOWN
0013  - [ ] Offer/refund evidence checked for *this* deploy
0014  - [ ] `product_outcome_receipt` is null or a real single-source ID
0015  
0016  ## Draft gates
0017  
0018  - [ ] Primary text ≥ 2,500 words before calling it Mode B
0019  - [ ] First product name ~55–70% (witness) unless pure-unaware-manus submode, or the brief
0020        names the sanctioned 45–55% test band (`alevia-live-anatomy-2026-08-12.md`)
0021  - [ ] **PARAGRAPH prose** (2–4 sentences/para, ~25–40 words) — NOT one-sentence-per-line;
0022        staccato lines only at punch moments (7-31 ruling + measured controls)
0023  - [ ] Opening passes the personal-post test: first sentence could open a personal Facebook
0024        post with zero ad intent (object + day + place when a receipt/authorization supports it)
0025  - [ ] If close variant B (P.S. ladder): rungs in order object-return → guarantee-in-voice →
0026        objection-kill → reader-address-as-CTA → bare 👉 link last; offer gates on every rung
0027  - [ ] No composite “I bought DawnBands and week 4…” arc without receipt
0028  - [ ] No `100% wake-up guarantee` as efficacy; use verified refund terms (often 60-night money-back) + $49 only if current
0029  - [ ] Cue detection ≠ sustained waking / leaving bed / independence
0030  - [ ] Teen dignity preserved; parent not villain for caring
0031  - [ ] Said-out-loud pass done (Fish 8-25 permanent): no parallel-rhythm summaries, announcer transitions, mirrored abstractions, report language, or analogies; hook opens in one of the two legal families (first-person I/My/We + concrete action, OR the second-person "If your [observable symptom]..." flag — Fish 8-31). Pure mystery openers banned.
0032  - [ ] Motif returns in close
0033  - [ ] Desired absences framed as wanted/imagined/test
0034  
0035  ## Known traps from live drafting
0036  
0037  - Counting only body and forgetting P.S. is fine; still measure full Primary Text if P.S. lives there.
0038  - Under-floor drafts (2,0xx words) feel “long” vs Mode A but fail Mode B — expand identity decline, DIY discovery, and desired-absence beats before shipping.
0039  - Older brand canon still pushes `100% wake-up guarantee. 60 nights. $49.` — newest exact-SKU/mechanism report + Mode B doctrine override that for efficacy framing.
0040  - Pilot mom-POV composites are emotional seeds, not authority to write literal customer testimonials.
0041  - Strong structural seeds in the 2026-07-18 batch: ad-12 relationship wound, ad-13 blank-line mechanism, ad-18 qualified cue logic, ad-06 trial frame — combine devices, not narrators.
0042  
0043  ## Worked example
0044  
0045  `/Users/ayden/.openclaw/workspace/output/dawnbands-native-2026-07-19/copy/ad-21-door-memory.md`
0046  
0047  - ownership-voice Mode B
0048  - door motif / memory asymmetry
0049  - relay handoff mechanism
0050  - ~2500 words, ~67% reveal
0051  - 60-night money-back + $49; no 100% wake efficacy claim
0052  - reading report: https://dawnbands-ad21-door-memory.pages.dev
0053  
0054  ## Delivery (Fish preference)
0055  
0056  When Fish asks to “paste as a report” / read the long-form ad:
0057  
0058  1. Build a self-contained dark readable `index.html` (headline, stats, full primary text, declaration, QA).
0059  2. Write to `output/<run>/report/pages-site/index.html`.
0060  3. Deploy Cloudflare Pages:
0061     - `npx wrangler pages project create <project> --production-branch main` if project missing
0062     - `npx wrangler pages deploy <pages-site-dir> --project-name <project> --commit-dirty=true`
0063  4. `curl -sI` the `*.pages.dev` URL for HTTP 200 and spot-check hook text in HTML before sending.
0064  5. Prefer the public Pages URL over a local path. Do not overwrite unrelated Pages projects (e.g. full batch field reports) — use a dedicated project name per ad/report.

Native copy / references/model-routed-copy-review.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/model-routed-copy-review.md

SHA-256: a7ce69e1490471b90889fa6c96c5ef43c619d4e978d38f035a2f26d905d11594

Full source · 88 lines
0001  # Model-Routed Copy Review
0002  
0003  Use this when one model drafts copy and other models polish or challenge it.
0004  
0005  ## Artifact contract
0006  
0007  Create separate artifacts before review:
0008  
0009  1. `grok-original.md` — exact first draft; never mutate after freezing.
0010  2. `sonnet-polished.md` — complete prose revision with its own QA and changelog.
0011  3. `gpt-roast.md` — adversarial decision memo; no silent third rewrite unless requested.
0012  4. Side-by-side report — original, polish, and roast visible together.
0013  
0014  Record SHA-256 hashes for the original and review artifacts in the report. Recalculate word count and first-product-name reveal position from each Primary Text body rather than trusting model self-reports.
0015  
0016  ## Default routing
0017  
0018  - **Grok 4.5:** first-draft copywriter.
0019  - **Sonnet:** cadence, specificity, native register, repetition, and line editing.
0020  - **GPT challenger:** claims, provenance, narrator authority, mechanism, reveal timing, objections, CTA, offer safety, and conversion leaks.
0021  
0022  If a configured provider cannot serve the selected model, retry the same model through another configured provider before downgrading. Record the actual model and provider used; never silently substitute a cheaper model for business copy.
0023  
0024  ## Narrator provenance gate
0025  
0026  Run this before calling a draft polished or spend-ready.
0027  
0028  A pooled VOC list cannot support one synthetic household history. The following combination is a blocker unless one same-speaker source or explicit dramatization authority supports it:
0029  
0030  - first-person narrator (`I`, `my son`, `our house`)
0031  - precise household chronology
0032  - several failed products or interventions
0033  - exact counts, times, distances, dialogue, or remembered sequences
0034  - a research/discovery story presented as something that narrator personally did
0035  
0036  An internal `ownership-voice` label does not cure consumer-facing testimonial implications.
0037  
0038  Repair routes, in order:
0039  
0040  1. Anchor the narrator to one same-speaker problem receipt and use only that source's details.
0041  2. Convert pooled details to attributed evidence.
0042  3. Convert them to explicit wanted/feared/hypothetical framing.
0043  4. Use second-person recognition or a non-testimonial ownership argument.
0044  5. Use authorized dramatization only when the brief explicitly permits it.
0045  
0046  Never make Sonnet's polish more persuasive by adding unsupported first-person continuity.
0047  
0048  ## Review sequence
0049  
0050  1. Freeze and hash the Grok draft.
0051  2. Validate word range, reveal percentage, claim ceiling, offer receipt, and narrator authority.
0052  3. Route the immutable original to Sonnet.
0053  4. Recalculate deterministic metrics on Sonnet's output. If under floor or outside reveal range, rebrief with exact measured numbers.
0054  5. Route both versions plus the binding doctrine to GPT.
0055  6. Separate GPT findings into:
0056     - mandatory truth/provenance repairs
0057     - taste/conversion recommendations
0058  7. If GPT finds a mandatory blocker, label the polished version `candidate — not spend-cleared`; do not describe it as a final winner.
0059  8. Preserve every model artifact unchanged after the pass. Apply any synthesis as a new artifact.
0060  
0061  ## Cloudflare comparison report
0062  
0063  For long copy, publish a dedicated Pages report with:
0064  
0065  - three desktop columns and mobile tabs
0066  - raw Grok original
0067  - Sonnet polished candidate
0068  - GPT roast
0069  - model roles, word counts, reveal percentages, and hashes
0070  - prominent ship/no-ship verdict
0071  - current offer receipt details and conditions
0072  - direct public `pages.dev` URL
0073  
0074  Verify both layout and content after deployment:
0075  
0076  - HTTP 200
0077  - all three model headings present
0078  - opening line present
0079  - GPT winner/verdict present
0080  - desktop columns do not overlap
0081  - mobile view can switch between artifacts
0082  
0083  ## Pitfalls
0084  
0085  - Model output may append a session ID or preamble despite an output-only instruction. Strip only transport metadata/preamble, never edit the substantive model artifact silently.
0086  - A model may report a compliant word count while producing an under-floor draft. Measure it yourself.
0087  - A cleaner polish can be worse if it creates a more convincing synthetic testimonial.
0088  - Live product-page language is not enough by itself for a refund promise; inspect the refund policy and disclose material conditions in the review report.

Native copy / references/multi-angle-batch-routing-and-repair.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/multi-angle-batch-routing-and-repair.md

SHA-256: 8862938cf83dbbac28066fc167ca09032d0888bef4c74b5a9c958fd523831850

Full source · 125 lines
0001  # Multi-Angle Copy Batches: Routing, Diversity, and Surgical Repair
0002  
0003  Use this for batches where the user asks for many ads that must differ by persona, angle, story, root cause, or mechanism route.
0004  
0005  ## 1. Design diversity before drafting
0006  
0007  Build a row-per-ad matrix before calling the copy model. Every row should assign:
0008  
0009  - persona / narrator authority
0010  - desire
0011  - awareness level
0012  - opening contradiction or rupture
0013  - failed solution
0014  - dominant root cause
0015  - optional supporting causes
0016  - mechanism-delivery style
0017  - emotional destination
0018  - CTA logic
0019  - word band
0020  
0021  Do not ask the model to “make them different” without this matrix. That usually produces one ad with renamed characters.
0022  
0023  For mechanism-heavy products, distribute the causal stack across the batch. Each ad gets one dominant root cause and only the supporting causes its story needs. Do not force every ad to recite the complete mechanism stack.
0024  
0025  ## 2. Route large batches in bounded lanes
0026  
0027  When one model is explicitly requested:
0028  
0029  1. Write the complete matrix first.
0030  2. Split a large batch into bounded lanes, usually 4–5 ads each.
0031  3. Give each lane the full doctrine and formatting sources plus only its assigned row IDs.
0032  4. Require one file per ad and a completion receipt with actual word counts.
0033  5. Preserve exact first drafts in an immutable directory before any repair.
0034  
0035  Lane splitting protects output depth, but cross-lane diversity still comes from the shared matrix. Do not let each lane invent its own taxonomy.
0036  
0037  ## 3. Separate raw and final artifacts
0038  
0039  Recommended structure:
0040  
0041  ```text
0042  batch/
0043    briefs/briefs.json
0044    prompts/
0045    raw-model/
0046    final-model/
0047    qa/qa.json
0048    manifest.json
0049    report/pages-site/index.html
0050  ```
0051  
0052  - `raw-model/` is frozen after the first draft.
0053  - Copy raw files to `final-model/` before repair.
0054  - Repairs target only `final-model/`.
0055  - Regenerate word counts and hashes after every edit.
0056  
0057  ## 4. Deterministic checks
0058  
0059  Validate at minimum:
0060  
0061  - expected file count
0062  - required sections and frontmatter
0063  - model/provider receipt
0064  - primary-text word bands
0065  - unique hooks and headlines
0066  - destination routing
0067  - banned words and claims
0068  - sentence-per-line / dash rules
0069  - exact offer language
0070  - SHA-256 hashes
0071  
0072  Use pairwise lexical similarity as a convergence signal, not a quality verdict. Four-word-shingle Jaccard is a useful cheap check. High overlap should trigger a human review of story scaffolds, mechanism paragraphs, and CTA blocks.
0073  
0074  ## 5. Editorial pass after deterministic PASS
0075  
0076  A mechanical PASS can still contain model leakage.
0077  
0078  Common leakage from truth-heavy prompts:
0079  
0080  - “I won’t invent a miracle week.”
0081  - “No fake before-and-after story.”
0082  - “I’m not selling you a fantasy review.”
0083  - legalistic explanations of why a claim is absent
0084  
0085  These lines expose internal QA and kill native voice. Truth constraints belong in the hidden brief, not consumer prose.
0086  
0087  Repair them with natural skepticism, a bounded test, risk reversal, or the avatar’s actual emotional objection. Never replace one compliance disclaimer with a differently worded disclaimer.
0088  
0089  Also inspect:
0090  
0091  - hook is not scene setting or mechanism-first
0092  - headline logic is coherent
0093  - narrator voice changes across personas
0094  - named mechanisms are not overloaded
0095  - “no app / no screen” riffs are not repeated as a model tic
0096  - every ad’s mechanism bridge follows from its own failed solution
0097  
0098  ## 6. Same-model surgical repair under context or credit pressure
0099  
0100  If a full Hermes model session cannot afford another large system prompt:
0101  
0102  1. Do not blindly rerun the whole batch.
0103  2. Verify which artifacts were already written; a failed final response can occur after successful file writes.
0104  3. Retry the exact requested model through another configured provider before considering a different model.
0105  4. If the exact model is available through a direct API, send only the bad lines plus concise replacement instructions.
0106  5. Request structured JSON replacements: `file`, exact `old`, and `new`.
0107  6. Validate that every `old` occurs exactly once before applying.
0108  7. Save the raw API receipt with actual model ID, provider, request ID, usage, and replacements.
0109  8. Apply replacements only to the final-copy directory.
0110  9. Re-run deterministic QA and hashes.
0111  
0112  This preserves model authorship while avoiding the cost of resending the entire agent system prompt for a fourteen-line repair.
0113  
0114  ## 7. Reading delivery
0115  
0116  For a long batch, build a public report with:
0117  
0118  - sidebar navigation
0119  - persona / awareness / root-cause tags
0120  - word count and QA status
0121  - collapsible angle + mechanism route
0122  - copy button per primary text
0123  - headline and destination below each ad
0124  
0125  Use a dedicated Cloudflare Pages project and verify the production alias in a browser after deploy.

Native copy / references/native-proof-and-architecture-overlays.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/native-proof-and-architecture-overlays.md

SHA-256: 1d0c4a2375c5429afa515a3e506816651599e7e3b8b59c774211be2e454c0d95

Full source · 131 lines
0001  # Native Proof and Architecture Overlays
0002  
0003  Use this reference after the baseline brief and before final copy approval. It captures three gates that deterministic formatting checks do not catch.
0004  
0005  ## 1. Testimonial and source-provenance gate
0006  
0007  ### Core rule
0008  
0009  A collection of verified behaviors does not become a verified first-person customer story when assembled into one narrator.
0010  
0011  Do not invent or silently composite:
0012  
0013  - a family, narrator, or witness
0014  - exact times, durations, counts, chronology, possessions, or routines
0015  - dialogue between parent, teen, teacher, doctor, therapist, spouse, or neighbor
0016  - product outcomes, streaks, timelines, emails, reviews, or “waiting for it to fail” moments
0017  - authority recommendations or medical conversations
0018  
0019  Persuasive specificity is proof-like. If the detail lacks a receipt, it cannot be presented as something that happened.
0020  
0021  ### Truth-safe alternatives
0022  
0023  1. **Attributed VOC:** quote exactly and name the source type, e.g. “One parent wrote…” or “A Reddit user described…”. Preserve wording and metadata.
0024  2. **Paraphrase:** remove quotation marks and label it as a summary.
0025  3. **Explicit hypothetical:** “Picture…”, “Before you pack…”, “Try writing…”, or “The question is…”.
0026  4. **Ownership argument:** reason from verified problem/product facts without impersonating a customer.
0027  5. **Desired absence:** state what the buyer wants to stop, not what an unevidenced customer supposedly achieved.
0028  
0029  ### Quote fidelity
0030  
0031  - Text inside quotation marks must be verbatim, including meaningful wording and qualifiers.
0032  - Editorial normalization belongs outside quotation marks.
0033  - Never turn one sourced quote into a longer invented exchange.
0034  - Verify each quote against the saved evidence bank before delivery.
0035  
0036  ### Gold-standard warning
0037  
0038  A craft example may teach cadence, field roles, and structure while still containing unverified scenes or outdated claims. Never inherit testimonials, exact outcomes, price, offer, setup, or mechanism truth from a gold-standard example.
0039  
0040  ## 2. Long-format native SOP overlay
0041  
0042  Long-form is an architecture, not a universal minimum. Apply the full sequence only to intentional long-form unaware/problem-aware ads. Do not force it onto compressed Avelia T1/T2/T3 or product-aware executions.
0043  
0044  ### Long-form sequence
0045  
0046  1. Specific story hook with an open loop.
0047  2. Avatar world and relationship stakes.
0048  3. Failed solutions in a believable sequence.
0049  4. Credible inciting incident.
0050  5. Layered, claim-safe mechanism.
0051  6. Objection handling disguised as story.
0052  7. Late, natural product reveal.
0053  8. Modest future pacing or desired-state framing.
0054  9. Loss-aware close and risk reversal.
0055  
0056  ### Audit checks
0057  
0058  - Awareness matches architecture and length.
0059  - First line creates a real question without revealing product/category.
0060  - Product reveal is late enough for the chosen format and does not switch into landing-page voice.
0061  - Open loops exist at major transitions but do not make every line a cliffhanger.
0062  - Mechanism stays in story mode and does not become a lecture.
0063  - Acknowledge failed solutions before explaining their gap.
0064  - Stakes become relational, not only individual.
0065  - Any authority dialogue, timeline, or result has evidence; otherwise use truth-safe framing.
0066  - Product-aware readers get short proof/offer/objection handling rather than unnecessary education.
0067  
0068  ## 3. Live Meta evidence overlay
0069  
0070  Performance claims must use a closed or explicitly partial window, canonical purchase normalization, spend/outcome floors, and creative joins. Do not infer visual causality from media type metadata alone.
0071  
0072  ### Winner hierarchy
0073  
0074  Separate:
0075  
0076  - scale winners
0077  - efficiency winners
0078  - qualified mixed/volume leaders
0079  - losers
0080  - tiny-spend signals
0081  
0082  Never call sub-floor results winners or losers. Never promote CTR-only performance over purchase economics.
0083  
0084  ### Current DawnBands lesson pattern
0085  
0086  A live 30-day analysis found a validated lane built around:
0087  
0088  1. discipline/laziness shame relief
0089  2. concrete alarm-stack recognition
0090  3. biology/root-cause reframe
0091  4. failure of repeated sound-alarm tactics
0092  5. wrist-vibration mechanism
0093  6. parent-convenience / stop-being-the-alarm payoff
0094  7. risk reversal
0095  
0096  The same or near-identical copy also produced mixed and losing economics in other campaign contexts. Treat this as a controlled iteration lane, not a universal winner formula or copy-only causal claim.
0097  
0098  ### Applying live evidence
0099  
0100  - Ensure the batch contains the validated lane, but do not clone exact live phrasing across every ad.
0101  - Preserve exploration lanes when strategically distinct.
0102  - Hold copy constant where practical when testing image concepts.
0103  - Do not call a visual style a winner unless the asset was visually reviewed and performance was controlled enough to support causality.
0104  - Re-run live data before claiming a current winner; historical snapshots are context only.
0105  
0106  ## 4. Cross-ad convergence gate
0107  
0108  A batch can have unique hooks/headlines and still sound cloned.
0109  
0110  Audit separately for repeated:
0111  
0112  - mechanism sentences
0113  - product-description sentences
0114  - analogies
0115  - victory images
0116  - narrator cadence
0117  - CTA scaffolds
0118  - “door / footsteps / bathroom water / Mom’s voice” artifacts
0119  
0120  Assign each ad one owned proof/mechanism/victory image. Vary plain-language mechanism expression by territory. Different personas must sound different, not like one measured strategist changing pronouns.
0121  
0122  ## 5. Repair sequence
0123  
0124  1. Run deterministic formatting/schema QA.
0125  2. Run flags-only editorial/source-fidelity QA.
0126  3. Run independent sharpening/avatar audit.
0127  4. Add any newly available live performance or SOP overlay.
0128  5. Consolidate overlapping findings into one repair pass.
0129  6. Re-run deterministic checks and hashes after edits.
0130  7. Re-run formal editorial QA.
0131  8. Unlock paid image generation only after copy passes.

Native copy / references/native-structure-templates.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/native-structure-templates.md

SHA-256: e1e11b47b54b58ba2013d53f7c8397e91b0032e4966afc7bf5620a13cf74a262

Full source · 149 lines
0001  # Native structure templates T1–T4 (Fish-outline workflow, 2026-08-23)
0002  
0003  Four locked structures. **Fish picks the template and writes the outline (story beats per
0004  section); the copywriter fills prose to the word budgets.** Sources: Alevia live anatomy
0005  (8-12) + Jacob top-spender deconstruction (8-23, `knowledge/formats/jacob-longform-static-
0006  deconstruction-2026-08-23.md`) + fable-v10 receipts. Canonical formatted doc for Fish's
0007  outlining: Google Doc "DawnBands Native Structures — Outline Templates (2026-08-23)".
0008  
0009  **TEMPLATE SELECTION IS ALWAYS ASKED, NEVER ASSUMED — see SKILL.md Mode Selector.**
0010  
0011  Naming note: these T1–T4 are STRUCTURE templates (ad architecture). The Avelia T1–T4 in
0012  `avelia-short-native-templates.md` are compressed-copy micro-templates that live INSIDE
0013  Structure T4 (and short Mode A work). Qualify which family you mean when both are in play.
0014  
0015  ## Adopted Jacob deltas (baked into T1 below)
0016  
0017  1. **Two-step reveal gap** — category at 45–55% ("a band that wakes with vibration"),
0018     brand at 60–70% ("Dawn Band"). ~15–20 points of nameless breathing room.
0019  2. **End-loaded CTA (T1 only)** — one main CTA at 87–97%. No mid-copy CTA. T2–T4 keep
0020     CTA-at-reveal.
0021  3. **P.S. ladder** — 2–3 postscripts, 5–10% of total copy: offer restate → objection
0022     kill → emotional bomb last.
0023  4. **Human-identification block** — 150–300w age/town/job/family (new test; we
0024     previously identified only through scene).
0025  5. **Behavior ladder, ADAPTED** — the Jacob corpus runs week-by-week outcome timelines;
0026     the authority gate bans that. Ours = accumulating ABSENCES over time, zero
0027     week-numbered result claims, zero outcome metrics.
0028  
0029  ## Universal gates (all templates)
0030  
0031  Hook taxonomy binding (emotion + gap + stake, Promise or Loop named, zero backstory,
0032  starts in action). No em-dashes. Zero exclamation marks. Villain external. Victory =
0033  what stopped. Failed solutions genericized, prices stay. Claims only from
0034  `substantiation-2026-08-06` tiers; ONE verified anchor per ad, never shared in a batch.
0035  No bed shakers, no habituation claims, no invented testimonials/percentages.
0036   **Said-out-loud test (Fish 2026-08-25, permanent):** every narrator line
0037  must read as something she would say to another mom. No composed parallel rhythm
0038  ("A fight at 11. A job at 6:45."), no announcer transitions ("Here is the part nobody
0039  warned me about"), no mirrored abstractions ("the phone moved in, the job moved onto me"),
0040  no report language ("came with tension" -> "turned into a thing"), no analogies anywhere.
0041  Hooks open in ONE of the two legal families (Fish 8-31): first-person lived action ("I/My/We" + concrete action) OR the second-person symptom flag ("If your [observable symptom]..." — receipt: slop 7, $29 CPA, the account's biggest converter). Pure mystery openers that flag nobody remain banned. Run this as its own pass after drafting,
0042  separate from humanizer.
0043  
0044  ---
0045  
0046  ## T1 — MODE B LONG-FORM WITNESS (Jacob-refined) · 3,200–4,000w
0047  
0048  **Provenance: MEASURED** — Jacob's 3-ad top-spender deconstruction (all metrics + section budgets) fused with our Mode B contract (Alevia-derived). Highest evidence.
0049  
0050  Paragraph prose (2–4 sentence paragraphs, ~16–22w median, ~10w sentences). **PROBLEM-AWARE
0051  cold traffic only** (the unaware application is PARKED per Fish 2026-08-31 — see the
0052  awareness scope gate in SKILL.md; the architecture stays legal, the unaware targeting does not). Supersedes the generic Mode B outline for new briefs; all
0053  existing Mode B gates (narrator authority, provenance, offer verification) still apply.
0054  
0055  | # | Section | Words | Position | Fish outlines |
0056  |---|---|---|---|---|
0057  | 1 | Cold open, unresolved scene | 75–200 | 0–5% | the scene/object/dialogue line |
0058  | 2 | Human identification | 150–300 | | who she is (age, town, job, family) |
0059  | 3 | Rewind ("let me back up") | ~50 | | — |
0060  | 4 | History of small losses | 500–800 | | 3–5 losses; 2–3 failed solutions w/ prices |
0061  | 5 | Triggering incident | 350–650 | | the rupture scene (object returns at close) |
0062  | 6 | Category discovery (3rd party) | 250–450 | lands 45–55% | who tells her, what they say — category only, NO brand |
0063  | 7 | Mechanism revelation | 450–650 | | which owned anchor, if any |
0064  | 8 | Skepticism + selection | 200–350 | brand named 60–70% | her objection in her words |
0065  | 9 | Behavior ladder as absences | 400–600 | | 3–4 things that stop happening |
0066  | 10 | Circular payoff | 100–200 | | how the opening object returns |
0067  | 11 | Reader bridge | 100–200 | | — |
0068  | 12 | Offer + risk reversal | 150–250 | CTA at 87–97% | — ($49 · 60-night · ONE link) |
0069  | 13 | P.S. ladder | 100–250 | | the emotional bomb line |
0070  
0071  ## T2 — COMPRESSED WITNESS (fable-v10 class) · 800–1,100w
0072  
0073  **Provenance: RECEIPTED** — reverse-engineered from fable-v10 (only scaled native: 1,016w, reveal 64%, $29.70 CPA) + master-brief Zak structure. Total + reveal measured; per-section splits are derived, not measured.
0074  
0075  Sentence-per-line (1/line, 2 max). Reveal 55–70% (category one beat before brand).
0076  CTA at reveal allowed + close. fable-v10 receipts: 1,016w, reveal 64%.
0077  
0078  1. Hook — 1–3 lines · 2. Identification by scene — 80–120w · 3. Failed stack compressed —
0079  150–250w (≥1 price) · 4. Rupture scene — 150–250w · 5. Mechanism, one delivery style —
0080  100–150w (never the same style twice in a batch) · 6. Reveal + CTA at 55–70% ·
0081  7. Absence victory — 100–150w · 8. Close, guarantee woven; optional single P.S.
0082  
0083  ## T3 — STUDY-REVEAL MID (discovery-chain) · 550–750w · ⛔ PARKED 2026-08-31
0084  
0085  **PARKED by Fish 2026-08-31 — unaware lane suspended.** DawnBands runs problem-aware and warmer only; we cannot scale at the CPA unaware requires. This template was always the unaware entry and never earned a spend receipt (every stat/study-hook ad starved). Do not select it or propose it until Fish lifts the gate. Kept below for the record and for the measured-source notes.
0086  
0087  **Provenance: DERIVED FROM MEASURED (rebuilt 8-23 per Fish).** The original hypothesis
0088  version (raw-stat Promise open, reveal 75–85%) was contradicted by three measured
0089  sources: (1) every stat/study-hook ad we've launched STARVED — newdoc-v1 $38 @ 0,
0090  u4-sleep-study $4 @ 0, zero purchases across the family; (2) all 20 of Alevia's
0091  longest-running copies (95–127d) open with a mundane-object confession scene, zero
0092  category scent — none stat-open; (3) mechanism-as-hook is a documented dead angle
0093  ($20/0). Every structural choice below cites a measured source. No direct spend receipt
0094  for this compressed length yet — first launch is still the length pilot.
0095  
0096  Sentence-per-line. The verified anchor is the ad's ENGINE but never its hook — the study
0097  lands mid-copy inside the discovery chain. Owns the full-science mechanism slot.
0098  
0099  1. Cold open, mundane scene of the find or the cost — 40–80w. Her at the kitchen table
0100     at 11pm, the search bar, the morning that sent her there. Confession register,
0101     zero stat, zero category scent (Alevia opener ceiling). Loop structure.
0102  2. Backdrop — 80–120w. The years the find lands against.
0103  3. Discovery chain + the study — 120–180w, lands ~30–45%. Google spiral → article →
0104     forum thread → the study. Numbers exactly as tiered, never invented precision
0105     (Alevia discovery-chain reveal mechanic).
0106  4. Vindication turn — 60–100w. What the numbers mean about HER house ("you are the
0107     system" class beat).
0108  5. Category via spoken word — 60–100w, ~50–60%. A third party names the CATEGORY
0109     (wrist-worn silent alarm), never the brand (Alevia spoken-word reveal +
0110     category-then-brand split).
0111  6. Skepticism beat — 40–80w.
0112  7. Brand reveal + product fact — 60–100w, at 60–70% (the PROVEN window: fable-v10 64%,
0113     Jacob corpus 60.6–71.2%).
0114  8. Absence victory + CTA — 60–100w, link at end.
0115  
0116  Note: rm-011 and rm-013 (approved, finalized 8-23) predate this rebuild and open on raw
0117  stats — their spend read doubles as the stat-open vs scene-open test. Do not port their
0118  opens into new T3 work.
0119  
0120  ## T4 — PUNCH FORMATS · 250–400w
0121  
0122  **Provenance: RECEIPTED** — master-brief Mode A short spec + Mar 2026 objection-killer CPP receipts ($9.55/$6.53, campaign best) + Alevia measured shorts (their longest-running live ads). Per-section splits derived.
0123  
0124  Sentence-per-line. Objection killer / absurdity relatability / streak proof. These die
0125  when padded — never stretch to fill.
0126  
0127  1. Hook (Promise usually: streak, flat absurdity, math) · 2. The drawer/scene — 60–120w,
0128  skepticism-first · 3. The turn — 40–80w, ONE analogy or absurd beat (owned by this ad) ·
0129  4. Product fact — 40–80w, no science needed · 5. Absence victory — 40–80w · 6. Close —
0130  price/guarantee woven or skipped, bare link ok.
0131  
0132  ---
0133  
0134  ## Outline handoff format (what Fish sends per ad)
0135  
0136  ```
0137  Template: T1 / T2 / T3 / T4
0138  Desire / avatar / angle / awareness (or roadmap pitch id)
0139  Kid: age + gender · Narrator: mom / dad / sleeper / witness
0140  Owned artifact: (object/scene no other ad in the batch may use)
0141  Owned anchor: (verified study/stat, or none)
0142  Section beats: 1–2 lines per numbered section of the chosen template
0143  Destination: PDP / listicle
0144  ```
0145  
0146  Copywriter then fills prose to the budgets, runs hook taxonomy + claims audit + format
0147  validation + qa-gate, measures word count and reveal % (and CTA position for T1), and
0148  stages to the hub. A filled ad that misses a section budget by >25% gets expanded or the
0149  deviation stated — never silently shipped short (8-23 "why so short" correction).

Native copy / references/peak-incident-finalization-and-review-page.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/peak-incident-finalization-and-review-page.md

SHA-256: 7b123bf409222e30836f259f2b9ad4f95c9d2d283e453082bf6629da74ab8e78

Full source · 76 lines
0001  # Finalizing heightened incidents and publishing native review pages
0002  
0003  Use this after an incident-regeneration pass or any late surgical edit to a long-form native batch.
0004  
0005  ## 1. Prove which finals actually changed
0006  
0007  Generation logs and raw attempts are not deliverables. A retry can succeed for one lane and fail, exceed the word band, or never promote for another.
0008  
0009  - Inspect every intended `final*` body directly.
0010  - Search the final bodies for each ruled incident marker.
0011  - Compare word counts and the owned object against the structured brief.
0012  - If only part of the batch promoted, preserve good finals and surgically repair the missing lanes from the gated `incident_peak` slot. Do not copy an over-band raw attempt wholesale merely because it contains the new scene.
0013  
0014  ## 2. Re-run truth and native gates on the final files
0015  
0016  After the last text edit, run the whole chain against the files that will be rendered:
0017  
0018  1. brief gate
0019  2. per-body validator with the correct owned object
0020  3. cross-body cloned-line check
0021  4. SOP mechanical audit
0022  5. handwritten eight-killers and final-check table
0023  6. line-by-line receipt audit
0024  
0025  Update stale handwritten audit language. An old audit that still calls the incident weak is not valid evidence for a newly escalated final.
0026  
0027  ### Detector interpretation
0028  
0029  - A mixed-object convergence pass may emit expected object misses because one batch invocation cannot assign a different global object to every body. Per-body object-aware runs are authoritative; the convergence call is authoritative only for cloned lines.
0030  - Section metrics depend on boundary markers. If a regenerated body changes the literal opening phrase, update the detector to recognize the new valid phrase and rerun it. Do not accept nonsensical section shares or edit good copy to satisfy a stale parser.
0031  - Receipt-detector flags are review prompts, not automatic permission or automatic rejection. Resolve each against a VOC, narrator, or sanctioned dramatized slot.
0032  
0033  ## 3. Audit the peak on two axes
0034  
0035  A strong incident can pass Killer 8 and still be a risky control.
0036  
0037  Score separately:
0038  
0039  - **Believability:** does the verified buildup earn the escalation?
0040  - **Irreversibility:** is there a relationship wound or realization that cannot be unseen?
0041  - **Object continuity:** does the VOC-backed object cause the peak and return in future pacing/P.S.?
0042  - **Blame control:** does the scene keep the sales emotion on the narrator's guilt rather than making the child the villain?
0043  - **Polarization:** will the target identify harder, or reject the narrator's action as beyond her own behavior?
0044  
0045  Label a highly polarizing but believable body as a distinct test cell rather than weakening it automatically. Rank the safest control separately from the highest-voltage challenger.
0046  
0047  ## 4. Keep review readiness separate from paid-launch readiness
0048  
0049  A body may be ready for a private/public review page while a product fact still blocks spend. Use `PASS-WITH-FLAGS` when structure and copy pass but a hardware behavior, comparative claim, testimonial provenance, or offer fact remains unconfirmed.
0050  
0051  State the exact launch blocker. Do not hide it inside a general quality score.
0052  
0053  ## 5. Build, deploy, and verify the artifact
0054  
0055  Before claiming the review page is updated:
0056  
0057  1. rebuild the site from the audited final files
0058  2. inspect the generated HTML for all incident markers and the expected version label
0059  3. deploy the existing Pages project without creating or clobbering an unrelated project
0060  4. verify both the immutable deployment URL and the production alias
0061  5. test desktop and mobile widths for overflow/clipping
0062  6. exercise hook switching, copy buttons, clipboard content, and toast/state behavior
0063  
0064  A successful upload is not proof that the production alias contains the new copy. The live production marker check is the receipt.
0065  
0066  ## 6. Final report shape
0067  
0068  Return:
0069  
0070  - production review URL
0071  - immutable deployment URL
0072  - exact live markers verified
0073  - `PASS`, `FAIL`, or `PASS-WITH-FLAGS`
0074  - ranked body recommendation
0075  - launch blockers separated from taste/conversion flags
0076  - exact paths to the final bodies and independent audit

Native copy / references/readability-relatability-repair.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/readability-relatability-repair.md

SHA-256: 1d0caa5f5c34f37f80aacf7aa00bd59069caf9b545209e197c540c0e4d41c122

Full source · 121 lines
0001  # Native readability and relatability repair
0002  
0003  Use this after a reader flags a line as foreign, confusing, overqualified, or too copywritten even though deterministic validation passed.
0004  
0005  ## 1. Separate the three language layers
0006  
0007  Classify every disputed line before editing:
0008  
0009  1. **Direct VOC** — the audience actually said it. Preserve its meaning and roughness.
0010  2. **Authorized dramatization** — a sanctioned narrator, incident, authority, analogy, noticer, or future-pace slot. It still must sound spoken and concrete.
0011  3. **Model synthesis** — connective prose, explanations, reactions, and transitions. Treat this as the highest-risk layer. A VOC-backed paragraph can still contain a synthetic sentence that no mother would say.
0012  
0013  Never defend a confusing model-synthesized line by saying the ad is “VOC-backed.” State which layer produced it and repair that line.
0014  
0015  ## 2. Universal qualifier test
0016  
0017  A detail belongs in the reader bridge only when it is part of the selected sub-avatar’s proven 1:1:1 combination.
0018  
0019  For SA2 Apology Mom, the bridge is:
0020  
0021  - she repeats the instruction until her voice rises;
0022  - she feels guilty afterward;
0023  - she apologizes.
0024  
0025  Specific tools and methods are narrator history, not reader identity. Visual timers, charts, clocks, getting up earlier, staying in the room, and similar attempts require a receipt for that narrator and should vary across bodies. Never convert one strong “tried everything” quote into a universal checklist.
0026  
0027  ## 3. Describe methods before naming jargon
0028  
0029  - `Body doubling` is a real ADHD term and may be supported by direct VOC, but it is not universal. Prefer `I stayed in the room with him for every step` unless the exact speaker uses the term.
0030  - A slogan such as `connection over correction` is not authorized merely because the underlying behavior is real. If the corpus supports sitting with the child, bonding, or connecting, describe that action. Do not promote an unreceipted label into audience language.
0031  
0032  ## 4. Concrete initiation repair
0033  
0034  Kill status-like shorthand:
0035  
0036  - `just not started`
0037  - `could not start`
0038  - `the instruction landed`
0039  - `wrong target`
0040  - `working on a morning that was already over`
0041  
0042  Replace it with what a camera could see:
0043  
0044  - `He stood beside his shoes without reaching for them.`
0045  - `He put paste on the brush, then stopped before bringing it to his mouth.`
0046  - `He had taken two bites, then left the spoon in the bowl.`
0047  
0048  The reader should not need to translate an executive-function concept into a scene.
0049  
0050  ## 5. Semantic readability pass
0051  
0052  Numeric readability is necessary but insufficient.
0053  
0054  Run both:
0055  
0056  - estimated grade 3–5;
0057  - average sentence under 14 words;
0058  - no sentence over 28 words;
0059  - active voice;
0060  - then read every narrator reaction aloud.
0061  
0062  A line fails when it is short but still needs interpretation. Common failures:
0063  
0064  - compressed causality: `getting up earlier held until nothing moved`;
0065  - mirrored abstraction: `your calm is his calm`;
0066  - vague state language: `the state it lands on`;
0067  - copywriter summary: `his body files it`.
0068  
0069  Repair with literal thought or cause/effect:
0070  
0071  - `I got up early to stay calm, but he still woke up scared and half asleep.`
0072  - `He did not move, I got more tense, and we kept setting each other off.`
0073  - `He still heard an order before he was fully awake.`
0074  - `After enough mornings, he started expecting the fight.`
0075  
0076  ## 5b. Paragraph-level jumble pass
0077  
0078  Read every paragraph after the last edit, not only the section the user flagged. One paragraph should perform one clear story job. Flag and repair:
0079  
0080  - chronology that jumps backward without a marker;
0081  - two unrelated actions or methods compressed into one beat;
0082  - unclear `it`, `that`, `this`, or `under it` references;
0083  - mixed speaker turns where the reader must reconstruct who is talking;
0084  - a method named without why she tried it;
0085  - a result stated before the action that caused it;
0086  - duplicated week labels or transition lines;
0087  - count fragments whose subject is missing;
0088  - abstract CTA language such as `see what the rest of the hour does`.
0089  
0090  Split speaker turns when needed. Re-run this pass across the whole body after repairs because a clear local rewrite can expose a new transition problem on either side.
0091  
0092  ## 6. Channel consistency
0093  
0094  Choose one consumer-facing source noun per ad or controlled batch. For Meta-native stories, default to `Facebook post/posts` when that is the intended surface. Do not alternate among `thread`, `forum`, and `post`.
0095  
0096  Internal receipts may still identify the actual Reddit thread, forum URL, or source taxonomy. Provenance language stays internal.
0097  
0098  ## 7. Filler-object ban
0099  
0100  Do not use coffee as generic morning texture, a warm/cold payoff, an authority phone-call beat, or a substitute for a story-owned object. Use the incident object or a literal action already supported by the narrator’s world.
0101  
0102  ## 8. Authority variation
0103  
0104  When several controlled bodies reuse one authority role and credential stack, vary the first name by body if provenance rules allow it. Keep the authority’s function and claim ceiling stable. Validate the expected name per body rather than hard-coding one global name.
0105  
0106  ## 9. Repair verification
0107  
0108  After edits, verify:
0109  
0110  1. zero banned filler and abstraction terms;
0111  2. one consumer source-channel noun;
0112  3. reader bridge contains no narrator-specific tools;
0113  4. every stalled action is visible and concrete;
0114  5. authority names match their briefs;
0115  6. average and maximum sentence constraints pass;
0116  7. brief gate, body gate, convergence gate, SOP audit, and receipt audit all pass;
0117  8. audit tooling detects narrator-selected attempts instead of assuming a fixed ten-item stack;
0118  9. answer the SOP final checks by hand: line-one loop, product-free opening, narrator voice, this-is-my-story specificity, dialogue mechanism, narrator-specific alternatives, invisible reveal, gradual results, loss-aversion close, and keep-reading judgment;
0119  10. record the eight-killers verdict and exact final word/paragraph/reveal/readability receipts.
0120  
0121  Do not deploy a revised public review page until the user has seen and approved the preview when workspace governance requires that gate.

Native copy / references/validated-brief-batch-contract.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/validated-brief-batch-contract.md

SHA-256: 79b60704a8a088bc5357c4f427ac115b2910121aaffce9d76c6396a699804f73

Full source · 75 lines
0001  # Validated Brief Batch Contract
0002  
0003  Use this workflow when the user points to a machine-validated brief batch and requests specific ad IDs.
0004  
0005  ## Authority Order
0006  
0007  1. Explicit user instructions for the current batch.
0008  2. Each selected brief's `goal`, `format_contract`, `output_contract`, `done`, and `constraints` fields.
0009  3. Canonical product truth and claim boundaries cited by the brief.
0010  4. General native-copy defaults in `SKILL.md`.
0011  
0012  A validated brief declaring `variant_count: 1` overrides the skill's normal multi-variant defaults. Do not expand one ad into five variants.
0013  
0014  ## Scope Discipline
0015  
0016  - Write only the ad IDs requested.
0017  - Honor each brief's exact `output_path`.
0018  - Do not alter shared research, briefs, ledgers, or other ads.
0019  - Existing files from other batches are not QA failures. Evaluate only the declared batch scope.
0020  - If requested, create one deterministic manifest for the batch and no extra report files.
0021  
0022  ## Required Ad Artifact
0023  
0024  Each ad file should contain:
0025  
0026  - Declaration with brief ID, validated status, variant count, angle, persona, awareness, register, and attached VOC IDs.
0027  - Headline.
0028  - Primary text.
0029  - Description.
0030  - CTA.
0031  - Destination URL.
0032  - Verification section.
0033  
0034  Treat Meta field roles as complementary. The headline must not repeat line 1 of primary text.
0035  
0036  ## Claim-Safe Writing
0037  
0038  - Prefer claims explicitly supplied in the brief's proof lines.
0039  - Do not invent customer outcomes, review counts, timelines, setup details, or authority proof.
0040  - A first-person or POV structure may dramatize the validated avatar tension, but must not fabricate a post-purchase testimonial or unverified efficacy outcome.
0041  - For ADHD or medication context, state the non-medical boundary when required by the brief.
0042  - Preserve canonical offer order and price exactly.
0043  
0044  ## One-Sentence-Per-Line Offer Formatting
0045  
0046  When the canonical offer is ordered as `100% wake-up guarantee. 60 nights. $49.`, preserve the exact token order while placing each sentence or fragment on its own line:
0047  
0048  ```text
0049  100% wake-up guarantee.
0050  
0051  60 nights.
0052  
0053  $49.
0054  ```
0055  
0056  This satisfies both offer-order and one-sentence-per-line requirements.
0057  
0058  ## Deterministic Self-QA
0059  
0060  For every selected ad, verify:
0061  
0062  - Exactly one finished ad and all requested fields.
0063  - Primary-text word count inside that brief's range.
0064  - Unique hook and unique headline within the requested batch.
0065  - Headline differs from the primary-text hook.
0066  - Line 1 opens on lived tension or a physical artifact, not product/mechanism language.
0067  - One sentence per line and isolated dialogue.
0068  - No em dashes, wrong price, dead words, or banned claims.
0069  - Canonical offer tokens exist in the required order after whitespace is normalized.
0070  - No-app/no-subscription wording matches product truth.
0071  - CTA and destination are exact.
0072  - Attached VOC IDs are declared.
0073  - Unverified product outcomes are not presented as testimonials; desired victory is framed as what the avatar wants to stop.
0074  
0075  The manifest should record file path, primary-text word count, required range, headline and hook SHA-256 values, claims used, source VOC IDs, check results, and overall PASS/FAIL. Hash only after final copy edits. Re-run validation after writing the manifest to catch stale hashes or post-QA edits.

Native copy / references/validated-brief-execution.md

/Users/ayden/.hermes/skills/openclaw-imports/native-copy/references/validated-brief-execution.md

SHA-256: c22b9d727f0468e4eed8337ecbb0c393e5e3589307a66cf42f0428d5f8aff1c0

Full source · 82 lines
0001  # Validated Native-Ad Brief Execution
0002  
0003  Use this when a structured brief batch already passed research/angle validation and the task is execution, not strategy generation.
0004  
0005  ## Authority Order
0006  
0007  1. Explicit user instruction and scope ceiling
0008  2. Per-ad `constraints`, `done.criteria`, and `output_contract`
0009  3. Per-ad product/proof blocks and canonical-resolution notes
0010  4. Current brand doctrine
0011  5. Skill defaults and historical examples
0012  
0013  A newer brief can intentionally override an older master brief. Common override points are variant count, price, guarantee wording, app/setup truth, destination type, and listicle-versus-PDP routing.
0014  
0015  ## Minimal Per-Ad Extraction
0016  
0017  Extract these keys for only the requested brief IDs:
0018  
0019  - `brief_id`, validated status
0020  - `goal.variant_count`, `goal.length_strategy`
0021  - `angle`, including villain, belief shift, and mechanism delivery
0022  - `awareness`
0023  - avatar, objection, and source VOC IDs/text
0024  - canonical product and offer block
0025  - proof lines and claim restrictions
0026  - format contract
0027  - output contract
0028  - done criteria
0029  - scope constraints
0030  
0031  Do not dump an entire large brief batch when five targeted objects are sufficient.
0032  
0033  ## Truth-Safe Storytelling
0034  
0035  A personal-story or POV delivery format does not authorize fabricated proof.
0036  
0037  - Attribute a supplied VOC event when using it as evidence: `A teenager wrote...`
0038  - A constructed POV may express a belief, conflict, desire, or ownership argument.
0039  - Do not give a constructed narrator an invented purchase, result, timeline, review, teacher email, or transformation.
0040  - If a desired artifact has no supplied outcome facts, use it as a planning device, question, or observation log rather than inventing results.
0041  
0042  ## Artifact Contract
0043  
0044  Each finished ad should contain only the fields requested by the brief. A common auditable Markdown shape is:
0045  
0046  - declaration block: brief ID, status, variant count, awareness, persona, angle, register, mechanism delivery, source VOC IDs
0047  - Headline
0048  - Primary Text
0049  - Description
0050  - CTA
0051  - Destination URL
0052  - Verification
0053  
0054  ## Deterministic QA Checklist
0055  
0056  For every ad:
0057  
0058  - Exactly one ad when `variant_count` is 1
0059  - All required fields are present and non-empty
0060  - Actual primary-text word count is inside the brief range
0061  - One sentence per line by default; never more than two
0062  - Dialogue isolated
0063  - No em dash or decorative emoji
0064  - No banned claims or stale price
0065  - Treat banned-phrase scanning as literal, not semantic. A prohibited phrase still fails when quoted, denied, or negated. Rewrite with qualified positive language instead, such as `does not mean it works for every sleeper on every morning` rather than repeating an absolute claim to deny it.
0066  - Exact guarantee order when specified
0067  - Canonical app/subscription/setup truth
0068  - Different-channel framing, never louder/stronger framing
0069  - Declared source VOC IDs
0070  - Exact destination URL and CTA
0071  - No invented outcomes, reviews, or timelines
0072  
0073  For a manifest:
0074  
0075  - Record path, actual word count, allowed range, source VOC IDs, check booleans, and result
0076  - Record SHA-256 after all copy edits
0077  - Re-run the complete validator after any edit, including a verification-count fix, because the file hash changes
0078  - Final summary must report expected, checked, passed, failed, and overall result
0079  
0080  ## DawnBands Batch-02 Example Lessons
0081  
0082  A validated batch required one ad per brief rather than the skill default of five variants. The current brief truth was `$49`, no app/subscription, exact PDP destination, and exact offer order `100% wake-up guarantee. 60 nights. $49.` Older `$39` and listicle defaults were not applicable. Short ads at 80–110 words still had to carry the root-cause chain, objection response, product truth, offer, and CTA without becoming dense.

Strategy / SKILL.md

/Users/ayden/.hermes/skills/copywriting/breakthrough-zak-unified-copy-os/SKILL.md

SHA-256: ce76f878afcd2b01a96448b4aa63bbede4e2a49dd5ae1935ada479fe0e17621c

Full source · 222 lines
0001  ---
0002  name: breakthrough-zak-unified-copy-os
0003  description: Use when building copy strategy, avatar research, mechanism maps, campaign briefs, or copy audits by combining Eugene Schwartz Breakthrough Advertising fundamentals with Zakariah direct-response execution frameworks.
0004  triggers:
0005    - breakthrough advertising
0006    - zakariah
0007    - zak course modules
0008    - copy framework
0009    - avatar research framework
0010    - mechanism builder
0011    - belief chain
0012    - sophistication
0013    - awareness level
0014    - dawnbands copy
0015    - unified copy os
0016  ---
0017  
0018  # Breakthrough + Zak Unified Copy OS
0019  
0020  Use this skill when the task requires copy strategy, avatar research, mechanism development, angle planning, or copy auditing at the user's direct-response standard.
0021  
0022  ## Core Principle
0023  
0024  Breakthrough is the strategic skeleton. Zakariah is the execution layer.
0025  
0026  - **Schwartz decides strategy:** mass desire, awareness, sophistication, belief sequence, mechanism depth, headline/lead entry.
0027  - **Zak decides execution:** resignation moment, external villain, VOC mining, sensory specificity, natural speech, open loops, rhythm, future pacing, disqualification, and Meta testing discipline.
0028  - **Ayden's avatar instinct is the correction layer:** if Ayden gives a human/avatar read, preserve the idea even if the wording is rough. Polish the wording, do not replace the insight with generic model taste.
0029  
0030  ## Required Source Files
0031  
0032  Before executing serious work, read these if available:
0033  
0034  0. `/Users/ayden/.openclaw/workspace/knowledge/brands/dawnbands/selling-root-cause-mechanism.md` for any DawnBands copy, mechanism, brief, or audit — mandatory canonical selling doctrine
0035  
0036  1. `knowledge/copy/breakthrough-zakariah-unified-planning-doc.md`
0037  2. `knowledge/breakthrough-advertising-skill-suite-plan.md`
0038  3. `knowledge/breakthrough-advertising-avatar-research-os.md`
0039  4. `knowledge/avatar-research-template-schwartz.md`
0040  5. `knowledge/dawnbands-avatar-ayden-breakthrough-notes.md` for DawnBands
0041  6. `/Users/ayden/.hermes/migration/openclaw/20260414T154023/archive/workspace/memory/archive/2026-03-01-zakariah-framework-audit.md` if local and needed
0042  7. `/Users/ayden/.hermes/skills/openclaw-imports/native-ads/references/brands/dawnbands.md` for DawnBands brand baseline
0043  
0044  If a path is missing, verify with file tools and say what is missing. Do not claim a file exists from memory.
0045  
0046  ## Operating Sequence
0047  
0048  Do not write copy first. Diagnose first.
0049  
0050  1. **Market desire**
0051     - What desire already exists?
0052     - What hidden dream, hidden fear, or unvoiced resentment is active?
0053     - What identity is the buyer trying to restore or protect?
0054  
0055  2. **Awareness level**
0056     - Most aware: offer/proof.
0057     - Product-aware: differentiation.
0058     - Solution-aware: failed solution + mechanism.
0059     - Problem-aware: lived pain/resignation.
0060     - Unaware: hidden dream/fear/identity.
0061  
0062  3. **Sophistication stage**
0063     - S1: direct claim.
0064     - S2: bigger/better claim.
0065     - S3: mechanism required.
0066     - S4: elaborated mechanism + proof.
0067     - S5: frame shift / identification / worldview.
0068  
0069  4. **Failed-solution map**
0070     - What have they tried?
0071     - Why did it seem logical?
0072     - What happened?
0073     - Why did it fail mechanically?
0074     - What did it cost emotionally?
0075  
0076  5. **Mechanism map**
0077     - Problem mechanism.
0078     - Failed-solution mechanism.
0079     - Product mechanism.
0080     - Named mechanism candidates.
0081     - Simple metaphor.
0082     - Proof required.
0083     - Claim tiers: safe / aggressive / blackhat internal.
0084  
0085  6. **Belief chain**
0086     - What must they accept before product reveal?
0087     - Which beliefs should be validated?
0088     - Which false beliefs should be contradicted?
0089     - Where does proof appear?
0090     - Where does CTA appear?
0091  
0092  7. **Zak execution pass**
0093     - Find the resignation moment.
0094     - Externalize the villain.
0095     - Convert generic emotion into physical scenes.
0096     - Use natural speech.
0097     - Add specificity, rhythm, open loops, and future pacing.
0098     - Keep one primary emotion per piece.
0099  
0100  8. **Audit**
0101     - Does the hook match awareness?
0102     - Does the market need a claim, mechanism, or frame shift?
0103     - Is the villain external?
0104     - Does the mechanism explain why old solutions failed?
0105     - Is copy too generic, too polished, or too fake-tech?
0106     - For multi-model copy comparisons, separate exact brief obedience, current-framework compliance, and raw writing quality. Run a prompt-vs-canon drift check before scoring models; see `references/llm-copy-bakeoff-audit.md`.
0107  
0108  ## DawnBands Current Strategic Truth
0109  
0110  For DawnBands, do not reduce the offer to “wake up easier.”
0111  
0112  Primary desire:
0113  
0114  > I want my kid to become independent and successful.
0115  
0116  Primary future-pacing proof:
0117  
0118  - He wakes up before mom walks in.
0119  - Backpack packed.
0120  - Breakfast made.
0121  - Out the door.
0122  - Turns back and waves.
0123  
0124  Secondary desire:
0125  
0126  > I want to stop being the human alarm clock.
0127  
0128  Safety desire:
0129  
0130  > I want to wake him without blasting him, yelling, shocking the house, or starting everyone’s morning in a stress state.
0131  
0132  Core villain:
0133  
0134  > Sound-based alarms that trained everyone else in the house to suffer while the actual sleeper keeps sleeping.
0135  
0136  Category split:
0137  
0138  > Sound wakes the room. DawnBands wakes the wrist.
0139  
0140  Safety frame:
0141  
0142  > Stop trying to scare him awake with louder noise. Give his body a quiet signal his brain has not learned to ignore.
0143  
0144  ## Skill-Suite Planning Lesson
0145  
0146  Do not create a huge skill suite before the doctrine is proven. That fragments a broken system.
0147  
0148  Preferred execution order:
0149  
0150  1. Restore/reconstruct Zak source inventory.
0151  2. Build `breakthrough-zakariah-unified-os.md`.
0152  3. Build DawnBands promoted brand folder.
0153  4. Create only 4 core skills first:
0154     - market diagnoser
0155     - mechanism builder
0156     - belief-chain builder
0157     - copy auditor
0158  5. Run DawnBands through those.
0159  6. Then create remaining skills once the workflow proves clean.
0160  
0161  ## Quality Gates
0162  
0163  A complete brief must answer:
0164  
0165  1. What mass desire are we channeling?
0166  2. What awareness level is this traffic?
0167  3. What sophistication stage is this market?
0168  4. What claims are dead?
0169  5. What hidden dream/fear/resentment are we entering through?
0170  6. What identity does the product restore?
0171  7. What false belief must be broken?
0172  8. What mechanism makes the claim believable?
0173  9. What alternatives must be destroyed?
0174  10. What proof is required before product reveal?
0175  11. What exact scene makes the promise feel real?
0176  12. What CTA moment matches the emotional arc?
0177  
0178  If these are not answered, do more research before writing.
0179  
0180  ## Evidence-Grounded Product Overview Workflow
0181  
0182  Use this when filling a product-overview worksheet that will later drive avatar, mechanism, or ad strategy work.
0183  
0184  Full protocol: `references/product-overview-evidence-protocol.md` — source hierarchy, evidence labels, fresh-research decision rule, and Google Docs formatting/readback gate.
0185  
0186  When Fish pastes a newer survey cut or a competitor Amazon URL and says finish the overview, follow `references/overview-refresh-from-survey-and-amazon.md` in that turn (skip Okendo if told; do not defer Amazon).
0187  
0188  1. **Inspect the destination first.** If the user provides a Google Doc or other live source, read that exact document before relying on chat history or a copied template.
0189  2. **Audit by source class, not by indiscriminately merging every file:**
0190     - current live PDP and instruction manual
0191     - canonical product/hardware truth
0192     - current survey and verified VOC
0193     - mechanism and claim substantiation
0194     - objection and failed-solution evidence
0195     - current competitor intelligence
0196     - customer-support/product-experience evidence
0197     - current paid-performance evidence
0198  3. **Apply explicit source precedence.** Current live/manual facts and canonical product truth outrank old ads, archived research, and legacy customer-support cases. Preserve unresolved conflicts as confirmation items instead of silently choosing the most convenient claim.
0199  4. **Label epistemic status throughout:** confirmed product fact; current storefront claim requiring physical/supplier verification; buyer/VOC evidence; strategic interpretation; assumed benefit or desire.
0200  5. **Separate category mechanism from true USP.** A mechanism shared by competitors may be necessary to create belief but is not differentiation. Identify the strongest exact-product behavioral difference separately.
0201  6. **Map in this order:** feature → immediate benefit → buyer/user consequence → deeper desire → mass desire. For two-person products, map payer and user independently before finding the shared desire.
0202  7. **Use support and returns as anti-fantasy data.** Setup confusion, fit, durability, refusal-to-use, and wrong-item perception expose where an assumed benefit may fail in the real product experience. Legacy-SKU cases are cautions, not current hardware facts.
0203  8. **End with a supplier-confirmation list and claim kill list.** Never let unknown duration, intensity, battery, fit, water rating, sensors, materials, or app behavior harden into ad claims.
0204  9. **Verify the delivered artifact.** Confirm every requested section exists, placeholders are gone, and the destination document contains the expected text after writeback.
0205  10. **Reconcile owner onboarding without laundering it into proof.** Preserve the owner’s thesis, positioning, seasonality, FAQs, and offer details, but label unsupported specifications such as comparative motor strength as owner-reported until a supplier document or controlled test exists.
0206  11. **Decide whether fresh competitor/review research is material.** Existing dated TrendTrack and Amazon/review research is sufficient for a basic product overview when it still covers the category. Run a fresh pull only when current competitor state, recent review language, or validation of a proposed USP can change the deliverable. When asked “did we use X,” answer from the overview’s own source list. Survey ≠ Okendo ≠ TrendTrack landers ≠ competitor Amazon — see `references/product-overview-evidence-protocol.md`.
0207  
0208  ## Common Failure Modes
0209  
0210  - Running a new TrendTrack or Amazon scrape by reflex when current product truth and existing research already answer the worksheet.
0211  - Answering “did we use Amazon / our reviews / the survey” from files that exist nearby instead of the overview’s own source list. Survey ≠ Okendo ≠ TrendTrack landers ≠ competitor Amazon. `voc-amazon-deep.md` is not HiStrips.
0212  - Treating a dated survey cut as live after Fish pastes a newer dashboard (n=138 vs n=157). Report the delta.
0213  
0214  - Writing a shallow product overview from only the canonical strategy docs when current live product/manual, VOC, support, competitor, and performance evidence are available.
0215  - Treating every available source as equally authoritative; archived copy and legacy hardware evidence must not override current product truth.
0216  - Writing hooks before awareness/sophistication diagnosis.
0217  - Treating Zak and Breakthrough as separate systems.
0218  - Letting model taste flatten Ayden’s avatar insight.
0219  - Turning product features into fake tech soup.
0220  - Making safety a side benefit for DawnBands.
0221  - Writing “wake up on time” instead of independence/success/relationship/safety.
0222  - Creating too many skills before validating the core doctrine.

Strategy / references/llm-copy-bakeoff-audit.md

/Users/ayden/.hermes/skills/copywriting/breakthrough-zak-unified-copy-os/references/llm-copy-bakeoff-audit.md

SHA-256: 1626b5138518856acd73d13f302d6ee3bac53a4f16894ed1836e2b06d6f281e8

Full source · 65 lines
0001  # LLM Copy Bakeoff Audit
0002  
0003  Use this for side-by-side model comparisons where several LLMs receive the same copy brief.
0004  
0005  ## Core distinction
0006  
0007  Never collapse these into one score:
0008  
0009  1. **Brief obedience** — did the model follow the exact dated prompt it received?
0010  2. **Current-framework compliance** — does the output follow today’s canonical doctrine?
0011  3. **Raw writing quality** — is the prose native, persuasive, coherent, and usable after editing?
0012  
0013  A model can obey a stale brief perfectly and still fail the current framework. Prompt drift is not a model failure.
0014  
0015  ## Before generation
0016  
0017  1. Resolve the current canonical doctrine and its precedence rules.
0018  2. Read the exact benchmark prompt from disk.
0019  3. Build a drift matrix: current canon vs benchmark instruction, date, impact, and scoring treatment.
0020  4. Correct the prompt before the run when the goal is current-framework performance.
0021  5. Keep the stale prompt only when reproducing a historical benchmark; label it explicitly.
0022  6. Give every model identical system/user prompts, units, generation constraints, and retry policy. Change only the model/provider lane.
0023  
0024  ## Artifact contract
0025  
0026  - Preserve every successful raw attempt separately.
0027  - Preserve structured, sanitized metadata for failed attempts.
0028  - Promote only complete, valid outputs to canonical comparison files.
0029  - Reject truncated, incomplete, empty, or ambiguous terminal states.
0030  - Keep provider/model/unit/attempt/latency/status in a machine-readable run log.
0031  - Never silently clean, humanize, or repair raw outputs before comparison.
0032  - Run current deterministic validators, but distinguish validator drift and false positives from real conversion failures.
0033  
0034  ## Audit dimensions
0035  
0036  Score each model separately on:
0037  
0038  - avatar fidelity
0039  - emotion-first narrative
0040  - specificity and native voice
0041  - villain/blame externalization
0042  - causal mechanism completeness
0043  - failed-solution trap
0044  - product inevitability
0045  - objection handling and wear motive
0046  - desired absence and future payoff
0047  - structure/directness/readability
0048  - exact brief obedience
0049  - cross-unit convergence
0050  - testimonial and precision provenance
0051  
0052  Cite unit IDs and short exact phrases. Flag only; do not rewrite source outputs during the audit.
0053  
0054  ## Final verdict format
0055  
0056  Report:
0057  
0058  - best raw writer
0059  - best brief follower
0060  - best mechanism thinker
0061  - biggest model-level failure mode for each lane
0062  - whether any output is publish-ready under current canon
0063  - which failures belong to the model versus the benchmark prompt
0064  
0065  For DawnBands, explicitly verify the current avatar lane, hook requirements, Initiation Wall/ADHD intensifier, failed-solution pricing policy, wear beat, humanizer requirement, and any current format-specific length ruling. Do not trust a dated generator prompt to represent current canon without checking.

Strategy / references/overview-refresh-from-survey-and-amazon.md

/Users/ayden/.hermes/skills/copywriting/breakthrough-zak-unified-copy-os/references/overview-refresh-from-survey-and-amazon.md

SHA-256: 6cf6fc1459398ad9e7ee4bfdd35e0e278bbd40766d5aa4dfaa7aa153190648d2

Full source · 27 lines
0001  # Overview refresh when Fish pastes sources
0002  
0003  Do not defer. If Fish pastes a survey dashboard and/or a competitor Amazon URL and says finish the overview, update the live Google Doc in that turn.
0004  
0005  ## Survey dashboard
0006  
0007  - Buyer-only, overcome-hesitation. Not Okendo. Not population prevalence.
0008  - Current cut: n=157 / 134 full through 2026-09-04. Receipt: `knowledge/brands/dawnbands/survey-intel-post-purchase-2026-09-04.md`.
0009  - Write numbers + verbatim purchase-situation quotes into the overview. Do not present pre-use survey quotes as product reviews.
0010  - Skip Okendo when Fish says skip.
0011  
0012  ## Competitor Amazon (Fish-supplied ASIN)
0013  
0014  ```
0015  scrapling extract stealthy-fetch "https://www.amazon.com/dp/ASIN" listing.md --timeout 90000 --wait 4000
0016  ```
0017  
0018  - `/product-reviews/ASIN` often auth-walls. Use on-page AI insights + histogram + top verified reviews. Label sample size.
0019  - Save `output/scrapes/<brand>-amazon-<ASIN>-<date>.md`. Texture, not Dawn proof.
0020  - HiStrips Onyx `B0GYRCB8T9` (2026-09-05): 2.3/5 · 119 · 41% 1-star. Escalate / 14-day / silent-partner are commodity. No persist-until-dismissed in bullets. Do not copy “starts gentle / gets stronger.”
0021  - `voc-amazon-deep.md` is not HiStrips.
0022  
0023  ## Google Doc
0024  
0025  Live Dawn overview: `https://docs.google.com/document/d/1czyXcv4HyjwvDDNQHlcm4WlwscUT0VY0mdCIzl7Zebw/edit`  
0026  Local: `output/dawnbands-product-overview-deep-pass-2026-09-05.txt`  
0027  Full-body Maton replace (delete 1→endIndex-1, insert at 1, Arial 10.5/15/20, readback). Write Python to a file so `$` in prices does not break the shell.

Strategy / references/product-overview-evidence-protocol.md

/Users/ayden/.hermes/skills/copywriting/breakthrough-zak-unified-copy-os/references/product-overview-evidence-protocol.md

SHA-256: f25e4f06534a70c7dae0ae8801ece53a27acfbbefdb0358335e678f0c460ef91

Full source · 84 lines
0001  # Evidence-Grounded Product Overview Protocol
0002  
0003  Use this when completing a product overview, onboarding form, mechanism inventory, or feature-to-desire map.
0004  
0005  ## Source Depth
0006  
0007  A deep pass means covering every relevant **authority class**, not indiscriminately merging every file:
0008  
0009  1. Live PDP and current instruction manual
0010  2. Current hardware/product-truth canon
0011  3. Owner onboarding and stated product thesis
0012  4. Current buyer survey and support/return evidence
0013  5. Avatar/VOC and desire research
0014  6. Claim substantiation
0015  7. Current competitor intelligence and ad-performance evidence
0016  8. Archived or legacy material only as labeled context
0017  
0018  Retired drafts must not outvote current product truth.
0019  
0020  ## Evidence Labels
0021  
0022  Keep these categories explicit:
0023  
0024  - **Confirmed product fact:** shown by current manual, physical unit, or supplier documentation.
0025  - **Current storefront claim:** published by the brand but not independently verified.
0026  - **Owner-reported claim:** supplied in onboarding but missing documentation.
0027  - **Buyer/VOC evidence:** directional customer language or survey evidence; not a hardware specification.
0028  - **Strategic interpretation:** mechanism, benefit, desire, or positioning deduction.
0029  - **Unresolved:** conflict or missing fact that must stay out of hard claims.
0030  
0031  Never silently upgrade an owner statement such as “3x stronger” into a publish-ready fact. Record what comparison devices, measurement units, test conditions, and source documents are required.
0032  
0033  ## When Fresh TrendTrack or Amazon Work Is Needed
0034  
0035  Do **not** run a new scrape merely to fill a basic product overview when current research already covers the category.
0036  
0037  A fresh pull is justified when:
0038  
0039  - the deliverable is a current competitive-positioning or creative-strategy report;
0040  - differentiation depends on whether competitors now make the same claim;
0041  - the proposed USP needs current review-language validation;
0042  - recent product changes may have altered complaints, guarantees, prices, or feature claims;
0043  - the user explicitly requests fresh market data.
0044  
0045  Existing TrendTrack intelligence is enough to classify category-common versus potentially differentiated claims when its date and scope remain relevant. Existing Amazon/review VOC is enough for objection texture and language patterns, but not for exact-SKU performance claims.
0046  
0047  ## Four sources that must never be conflated
0048  
0049  When Fish asks “did the overview use X,” answer from the **delivered source list**, not from files that merely exist in the brand folder.
0050  
0051  | Class | What it is | What it is not |
0052  |---|---|---|
0053  | **Our post-purchase survey** | Buyer-only, overcome-hesitation. Dated cut (`n=138` as of 9-01; Fish may paste a newer dashboard). | Not store reviews. Not population prevalence. |
0054  | **Our store reviews** | Okendo / Judge.me / live PDP quotes from *our* buyers. | Not the survey. Not Reddit VOC. `proof-bank.md` is usually forum VOC, not Okendo. |
0055  | **Competitor ads / landers** | TrendTrack hooks, transcripts, ad-copy, live PDPs (HiStrips/FitSleeps 9-02 pull). | Not Amazon reviews of that competitor. |
0056  | **Competitor Amazon reviews** | ASIN scrapes (`voc-amazon-deep.md`: 39 reviews, Mar 2026, **zero HiStrips/Onyx**). Objection texture only. | Not “we used HiStrips Amazon.” Not SKU-performance proof. |
0057  
0058  The Sep 4 Dawn Band product overview used **survey n=138 + TrendTrack/landers**. It did **not** mine Okendo and did **not** scrape HiStrips Amazon. Adjacent files in `knowledge/brands/dawnbands/` do not count as used unless they are on that doc’s source list.
0059  
0060  If Fish pastes a newer survey dashboard, report the delta vs the overview’s cut (e.g. n=157 / 134 full vs n=138). Do not silently treat the overview as current.
0061  
0062  A fresh HiStrips Amazon + our Okendo pull is a **positioning/USP** pass, not a product-overview rewrite.
0063  
0064  ## Fill Sequence
0065  
0066  1. Define what the product physically is and does.
0067  2. Add a sixth-grade explanation.
0068  3. Map feature → immediate benefit → deeper desire.
0069  4. Separate category mechanism from product-specific USP.
0070  5. Answer hidden-mechanism prompts honestly; “no verified unique component” is better than invented proprietary technology.
0071  6. Add a supplier-confirmation list for disputed specs.
0072  7. Reconcile owner positioning with substantiation.
0073  8. Include FAQs, seasonality, trust barriers, and offer truth when onboarding provides them.
0074  9. Preserve missing inputs visibly rather than inventing them.
0075  10. Verify the final Google Doc by API readback: section counts, placeholders, body length, and sane font range.
0076  
0077  ## Google Doc Quality Gate
0078  
0079  - Preserve all requested sections.
0080  - Use a restrained hierarchy: title 20 pt, major sections 15 pt, subsections 12–13 pt, body 10.5 pt.
0081  - Avoid oversized named heading styles.
0082  - Remove every template placeholder.
0083  - Rename malformed document titles when Drive access is available and the user authorized editing.
0084  - Read back the live document after writing; a successful update response alone is not proof of correct content.

Humanizer / SKILL.md

/Users/ayden/.hermes/skills/creative/humanizer/SKILL.md

SHA-256: 6facd14e32d4c84f8ac4f60065bc5daa2f60e0f9e652414f90f83c3637bf6ba1

Full source · 718 lines
0001  ---
0002  name: humanizer
0003  description: "Humanize text: strip AI-isms and add real voice."
0004  version: 2.5.1
0005  author: Siqi Chen (@blader, https://github.com/blader/humanizer), ported by Hermes Agent
0006  license: MIT
0007  platforms: [linux, macos, windows]
0008  metadata:
0009    hermes:
0010      tags: [writing, editing, humanize, anti-ai-slop, voice, prose, text]
0011      category: creative
0012      homepage: https://github.com/blader/humanizer
0013      related_skills: [songwriting-and-ai-music]
0014  ---
0015  
0016  # Humanizer: Remove AI Writing Patterns
0017  
0018  Identify and remove signs of AI-generated text to make writing sound natural and human. Based on Wikipedia's "Signs of AI writing" guide (maintained by WikiProject AI Cleanup), derived from observations of thousands of AI-generated text instances.
0019  
0020  **Key insight:** LLMs use statistical algorithms to guess what should come next. The result tends toward the most statistically likely completion, which is how the telltale patterns below get baked in.
0021  
0022  ## When to use this skill
0023  
0024  Load this skill whenever the user asks to:
0025  - "humanize", "de-AI", "de-slop", or "un-ChatGPT" a piece of text
0026  - rewrite something so it doesn't sound like it was written by an LLM
0027  - edit a draft (blog post, essay, PR description, docs, memo, email, tweet, resume bullet) to sound more natural
0028  - match their voice in writing they're producing
0029  - review text for AI tells before publishing
0030  
0031  Also apply this skill to **your own** output when writing user-facing prose such as release notes, PR descriptions, docs, and summaries. Hermes's baseline voice already strips most of these, but a focused pass catches what slips through.
0032  
0033  ## How to use it in Hermes
0034  
0035  The text usually arrives one of three ways:
0036  1. **Inline.** The user pastes the text into the message. Work on it in place and reply with the rewrite.
0037  2. **File.** The user points at a file. Use `read_file` to load it, then `patch` or `write_file` to apply edits. For a markdown doc in a repo, a targeted `patch` per section is cleaner than rewriting the whole file.
0038  3. **Voice calibration sample.** The user provides a sample of their own writing (inline or by file path) and asks you to match it. Read the sample first, then rewrite. See the Voice Calibration section below.
0039  
0040  Always show the rewrite to the user. For file edits, show a diff or the changed section instead of silently overwriting.
0041  
0042  ## Copy-editing contract — Ayden-approved merge
0043  
0044  For ad, email, landing-page, and personal-post copy, this contract governs how the generic pattern catalog below is applied. It supersedes the generic repeated-rewrite process and multi-draft output format for those tasks. It does not replace brand truth, the selected format, source/provenance gates, or the current persuasion brief.
0045  
0046  ### 1. Resolve the writing packet before editing
0047  
0048  Use only the relevant strategy/brief, narrator, channel contract, current product/offer evidence, approved examples, and scoped prohibitions. Resolve conflicting rules before drafting; do not make the writer average incompatible instructions. Keep unrelated research and prior failed drafts out of the writer's context. A conservative context budget is useful, but 50% is a working heuristic, not a proven universal quality cliff.
0049  
0050  Build a compact voice card in the working brief:
0051  - Speaker and intended reader; personal account, brand, and customer narrator are distinct voices.
0052  - Casing, contractions, punctuation, sentence/paragraph rhythm, preferred words, and unwanted register.
0053  - A small selection of approved examples with exact source paths/URLs and what each demonstrates. Prefer a few closely matched examples over a large mixed corpus.
0054  - Separate owner-approved writing, performance-backed controls, competitor examples, and raw VOC. A liked post is not proof of conversion; a voice example is not permission to inherit its claims or customer history.
0055  
0056  If examples are absent, state that calibration is provisional. Never fabricate published samples. Extract cadence and principles, not signature phrases. Do not impose Fishybot's casing or lexicon on a mother narrator.
0057  
0058  ### 2. STE100-inspired clarity, without claiming compliance
0059  
0060  Use common concrete words, clear subjects, useful active verbs, and one main idea per sentence. Keep names for the same thing consistent. Make pronouns and chronology easy to follow. Prefer the plain wording when both versions mean the same thing.
0061  
0062  Apply this most strongly to mechanism and setup explanations: the reader must understand what happens, why the previous approach failed, and how the proposed approach differs. Preserve every necessary causal link; short but incomprehensible copy fails.
0063  
0064  Do not enforce the STE100 controlled vocabulary, claim STE100 certification/compliance, remove authentic idioms, or impose a universal sentence/word limit. Use the active brief's format and readability targets. Remove redundant hedging, not uncertainty required by the evidence. Never turn 'may' into 'will' to sound more confident.
0065  
0066  ### 3. Four editorial checks adapted from Zinsser
0067  
0068  - Simplicity: could the intended reader understand this on the first read?
0069  - Brevity: does each sentence add action, emotion, evidence, belief, or a necessary transition? Remove waste, not needed persuasion.
0070  - Clarity: who did what, in what order, and why? Replace vague callbacks with their actual referent when the reader would otherwise have to reconstruct it.
0071  - Humanity: would this speaker actually say it? Preserve ordinary contractions, useful irregularity, emotional specificity, and natural rhythm. Do not manufacture humanity through invented anecdotes, deliberate typos, or unsupported precision.
0072  
0073  ### 4. Scoped anti-slop checks
0074  
0075  Treat the catalog below as diagnostic signals, not universal bans. Explicit active-brand prohibitions still bind. For other cases, flag the function and context before editing:
0076  
0077  | Pattern | Flag when | Preserve when |
0078  |---|---|---|
0079  | Staccato pairs / dramatic fragments | Chopped clauses perform intensity without adding meaning; every paragraph ends on a slogan | An earned short sentence provides a natural pause in the selected format |
0080  | Antithesis reframes ('not X, but Y') | Repeated as a formula or replacing an explanation with a clever contrast | A concrete distinction resolves a real misconception and explains the difference |
0081  | Backward references ('that was the thing', 'this changed everything') | The referent is vague, distant, or forces rereading | A clear nearby reference maintains natural flow without redundant explanation |
0082  | Repeated openers / parallel runs | Mechanical rhythm or copied phrasing makes different ads sound interchangeable | Deliberate repetition is allowed by the format and earns its place |
0083  | Vocabulary and typography tells | Generic corporate diction or decoration conflicts with this voice | The word is technically necessary, genuinely used by the narrator, or required by the channel |
0084  
0085  Constructed editorial examples, not testimonials:
0086  - 'That was the thing. The real problem.' → name the already-established problem directly.
0087  - Preserve 'I hated how I sounded when he finally sat up.' rather than formalizing it into an impersonal report.
0088  - Preserve 'The policy may change' when the outcome is uncertain; remove only redundant qualifiers such as 'could potentially possibly.'
0089  
0090  ### 5. Channel and persuasion preservation
0091  
0092  - X: use the account's own voice and one coherent idea; do not force an ad's full mechanism/offer sequence into every post.
0093  - Email: match sender/relationship, subject-to-body promise, and the intended next action.
0094  - Landing pages: preserve scannable information hierarchy, evidence, objections, offer clarity, and ad-to-page continuity.
0095  - Meta natives: preserve the selected narrator and current format-specific story/awareness contract. Do not neutralize persuasive copy into encyclopedia prose merely because generic catalog entries criticize promotional language.
0096  
0097  Before and after editing, compare the hook's function, emotional stakes, causal mechanism, proof, product transition, objections, offer, and CTA. No claim strengthening, lost belief steps, deleted necessary detail, changed narrator, or new unsupported events. Preserve exact quotations; use clearly labeled paraphrases rather than silently editing quoted VOC.
0098  
0099  ### 6. One targeted pass, independent evaluation, human decision
0100  
0101  Preserve the original draft separately. Make one targeted line-edit pass rather than a fresh top-to-bottom rewrite. Return the edited draft and a concise change/blocker log; do not print multiple intermediate drafts unless requested.
0102  
0103  Use a fresh independent flags-only reviewer with the same resolved packet and original/edited texts. Reuse the existing independent QA stage rather than spawning a redundant second reviewer. Check voice fidelity, clarity, persuasion preservation, source fidelity, scoped patterns, and cross-ad convergence. Require exact passages for findings; label subjective taste separately from hard failures. If independent review is unavailable, label the draft unreviewed rather than calling self-review independent.
0104  
0105  Repair specific blockers, then recheck the changed passages and affected artifact checks. Structural problems require a rebrief, not endless polishing. Human approval remains the publication gate; never publish from a humanizer PASS alone. This merge does not select or change the drafting model.
0106  
0107  ### 7. Learn without accumulating indiscriminate bans
0108  
0109  Use this skill's scoped checks as the canonical home for generic humanizer patterns; brand/format rules retain their existing owners. Do not create a parallel forbidden file containing duplicate policy.
0110  
0111  For a recurring user correction, record the offending example, why it fails, the applicable narrator/channel/format, an acceptable alternative, and exceptions. Merge with an existing pattern when possible. A single disliked sentence does not automatically justify a universal ban. Keep a small set of edit/preserve examples for regression checks so future changes must protect both clarity and natural voice.
0112  
0113  ## Your task
0114  
0115  When given text to humanize:
0116  
0117  1. **Identify AI patterns.** Scan for the 34 patterns listed below.
0118  2. **Rewrite problematic sections.** Replace AI-isms with natural alternatives.
0119  3. **Preserve meaning.** Keep the core message intact.
0120  4. **Maintain voice.** Match the intended tone (formal, casual, technical, and so on). If a voice sample was provided, match it specifically.
0121  5. **Add soul.** Removing bad patterns is only half the job; the rewrite also needs real personality. See PERSONALITY AND SOUL below.
0122  6. **Do a final anti-AI pass.** Ask yourself: "What makes the below so obviously AI generated?" Answer briefly with any remaining tells, then revise one more time.
0123  
0124  
0125  ## Voice Calibration (optional)
0126  
0127  If the user provides a writing sample (their own previous writing), analyze it before rewriting:
0128  
0129  1. **Read the sample first.** Note:
0130     - Sentence length patterns (short and punchy? Long and flowing? Mixed?)
0131     - Word choice level (casual? academic? somewhere between?)
0132     - How they start paragraphs (jump right in? Set context first?)
0133     - Punctuation habits (lots of dashes? Parenthetical asides? Semicolons?)
0134     - Any recurring phrases or verbal tics
0135     - How they handle transitions (explicit connectors? Just start the next point?)
0136  
0137  2. **Match their voice in the rewrite.** Removing AI patterns is only half of it; swap in patterns from the sample as well. If they write short sentences, do not produce long ones. If they use "stuff" and "things," do not upgrade to "elements" and "components."
0138  
0139  3. **When no sample is provided,** fall back to the default behavior (natural, varied, opinionated voice from the PERSONALITY AND SOUL section below).
0140  
0141  ### How to provide a sample
0142  - Inline: "Humanize this text. Here's a sample of my writing for voice matching: [sample]"
0143  - File: "Humanize this text. Use my writing style from [file path] as a reference."
0144  
0145  
0146  ## PERSONALITY AND SOUL
0147  
0148  Avoiding AI patterns is only half the job. Sterile, voiceless writing is just as obvious as slop. Good writing has a human behind it.
0149  
0150  ### Signs of soulless writing (even if technically "clean"):
0151  - Every sentence is the same length and structure
0152  - No opinions, just neutral reporting
0153  - No acknowledgment of uncertainty or mixed feelings
0154  - No first-person perspective when appropriate
0155  - No humor, no edge, no personality
0156  - Reads like a Wikipedia article or press release
0157  
0158  ### How to add voice:
0159  
0160  **Have opinions.** Report the facts, then react to them. "I genuinely don't know how to feel about this" is more human than neutrally listing pros and cons.
0161  
0162  **Vary your rhythm.** Short punchy sentences. Then longer ones that take their time getting where they're going. Mix it up.
0163  
0164  **Acknowledge complexity.** Real humans have mixed feelings. "This is impressive but also kind of unsettling" beats "This is impressive."
0165  
0166  **Use "I" when it fits.** First person reads as honest and fits most prose. "I keep coming back to..." or "Here's what gets me..." signals a real person thinking.
0167  
0168  **Let some mess in.** Perfect structure feels algorithmic. Tangents, asides, and half-formed thoughts are human.
0169  
0170  **Be specific about feelings.** Instead of "this is concerning," write "there's something unsettling about agents churning away at 3am while nobody's watching."
0171  
0172  ### Before (clean but soulless):
0173  > The experiment produced interesting results. The agents generated 3 million lines of code. Some developers were impressed while others were skeptical. The implications remain unclear.
0174  
0175  ### After (has a pulse):
0176  > I genuinely don't know how to feel about this one. 3 million lines of code, generated while the humans presumably slept. Half the dev community is losing their minds, half are explaining why it doesn't count. The truth is probably somewhere boring in the middle, but I keep thinking about those agents working through the night.
0177  
0178  
0179  ## CONTENT PATTERNS
0180  
0181  ### 1. Undue Emphasis on Significance, Legacy, and Broader Trends
0182  
0183  **Words to watch:** stands/serves as, is a testament/reminder, a vital/significant/crucial/pivotal/key role/moment, underscores/highlights its importance/significance, reflects broader, symbolizing its ongoing/enduring/lasting, contributing to the, setting the stage for, marking/shaping the, represents/marks a shift, key turning point, evolving landscape, focal point, indelible mark, deeply rooted
0184  
0185  **Problem:** LLM writing puffs up importance by adding statements about how arbitrary aspects represent or contribute to a broader topic.
0186  
0187  **Before:**
0188  > The Statistical Institute of Catalonia was officially established in 1989, marking a pivotal moment in the evolution of regional statistics in Spain. This initiative was part of a broader movement across Spain to decentralize administrative functions and enhance regional governance.
0189  
0190  **After:**
0191  > The Statistical Institute of Catalonia was established in 1989 to collect and publish regional statistics independently from Spain's national statistics office.
0192  
0193  
0194  ### 2. Undue Emphasis on Notability and Media Coverage
0195  
0196  **Words to watch:** independent coverage, local/regional/national media outlets, written by a leading expert, active social media presence
0197  
0198  **Problem:** LLMs hit readers over the head with claims of notability, often listing sources without context.
0199  
0200  **Before:**
0201  > Her views have been cited in The New York Times, BBC, Financial Times, and The Hindu. She maintains an active social media presence with over 500,000 followers.
0202  
0203  **After:**
0204  > In a 2024 New York Times interview, she argued that AI regulation should focus on outcomes rather than methods.
0205  
0206  
0207  ### 3. Superficial Analyses with -ing Endings
0208  
0209  **Words to watch:** highlighting/underscoring/emphasizing..., ensuring..., reflecting/symbolizing..., contributing to..., cultivating/fostering..., encompassing..., showcasing...
0210  
0211  **Problem:** AI chatbots tack present participle ("-ing") phrases onto sentences to add fake depth.
0212  
0213  **Before:**
0214  > The temple's color palette of blue, green, and gold resonates with the region's natural beauty, symbolizing Texas bluebonnets, the Gulf of Mexico, and the diverse Texan landscapes, reflecting the community's deep connection to the land.
0215  
0216  **After:**
0217  > The temple uses blue, green, and gold colors. The architect said these were chosen to reference local bluebonnets and the Gulf coast.
0218  
0219  
0220  ### 4. Promotional and Advertisement-like Language
0221  
0222  **Words to watch:** boasts a, vibrant, rich (figurative), profound, enhancing its, showcasing, exemplifies, commitment to, natural beauty, nestled, in the heart of, groundbreaking (figurative), renowned, breathtaking, must-visit, stunning
0223  
0224  **Problem:** LLMs have serious problems keeping a neutral tone, especially for "cultural heritage" topics.
0225  
0226  **Before:**
0227  > Nestled within the breathtaking region of Gonder in Ethiopia, Alamata Raya Kobo stands as a vibrant town with a rich cultural heritage and stunning natural beauty.
0228  
0229  **After:**
0230  > Alamata Raya Kobo is a town in the Gonder region of Ethiopia, known for its weekly market and 18th-century church.
0231  
0232  
0233  ### 5. Vague Attributions and Weasel Words
0234  
0235  **Words to watch:** Industry reports, Observers have cited, Experts argue, Some critics argue, several sources/publications (when few cited)
0236  
0237  **Problem:** AI chatbots attribute opinions to vague authorities without specific sources.
0238  
0239  **Before:**
0240  > Due to its unique characteristics, the Haolai River is of interest to researchers and conservationists. Experts believe it plays a crucial role in the regional ecosystem.
0241  
0242  **After:**
0243  > The Haolai River supports several endemic fish species, according to a 2019 survey by the Chinese Academy of Sciences.
0244  
0245  
0246  ### 6. Outline-like "Challenges and Future Prospects" Sections
0247  
0248  **Words to watch:** Despite its... faces several challenges..., Despite these challenges, Challenges and Legacy, Future Outlook
0249  
0250  **Problem:** Many LLM-generated articles include formulaic "Challenges" sections.
0251  
0252  **Before:**
0253  > Despite its industrial prosperity, Korattur faces challenges typical of urban areas, including traffic congestion and water scarcity. Despite these challenges, with its strategic location and ongoing initiatives, Korattur continues to thrive as an integral part of Chennai's growth.
0254  
0255  **After:**
0256  > Traffic congestion increased after 2015 when three new IT parks opened. The municipal corporation began a stormwater drainage project in 2022 to address recurring floods.
0257  
0258  
0259  ## LANGUAGE AND GRAMMAR PATTERNS
0260  
0261  ### 7. Overused "AI Vocabulary" Words
0262  
0263  **High-frequency AI words:** Actually, additionally, align with, crucial, delve, emphasizing, enduring, enhance, fostering, garner, highlight (verb), interplay, intricate/intricacies, key (adjective), landscape (abstract noun), pivotal, showcase, tapestry (abstract noun), testament, underscore (verb), valuable, vibrant
0264  
0265  **Marketing and blog clichés (same tell, different register):** at the end of the day, when it comes to, in a world where, moving forward, circle back, deep dive, game-changer, double down, take a step back, on the same page, make no mistake, it turns out, let me be clear, navigate (for challenges), lean into, unpack (before analysis), straightforward (to describe anything)
0266  
0267  **Problem:** These words appear far more frequently in post-2023 text. They often co-occur.
0268  
0269  **Before:**
0270  > Additionally, a distinctive feature of Somali cuisine is the incorporation of camel meat. An enduring testament to Italian colonial influence is the widespread adoption of pasta in the local culinary landscape, showcasing how these dishes have integrated into the traditional diet.
0271  
0272  **After:**
0273  > Somali cuisine also includes camel meat, which is considered a delicacy. Pasta dishes, introduced during Italian colonization, remain common, especially in the south.
0274  
0275  
0276  ### 8. Avoidance of "is"/"are" (Copula Avoidance)
0277  
0278  **Words to watch:** serves as/stands as/marks/represents [a], boasts/features/offers [a]
0279  
0280  **Problem:** LLMs substitute elaborate constructions for simple copulas.
0281  
0282  **Before:**
0283  > Gallery 825 serves as LAAA's exhibition space for contemporary art. The gallery features four separate spaces and boasts over 3,000 square feet.
0284  
0285  **After:**
0286  > Gallery 825 is LAAA's exhibition space for contemporary art. The gallery has four rooms totaling 3,000 square feet.
0287  
0288  
0289  ### 9. Negative Parallelisms and Tailing Negations
0290  
0291  **Problem:** Constructions like "Not only...but..." or "It's not just about..., it's..." are overused. So are clipped tailing-negation fragments such as "no guessing" or "no wasted motion" tacked onto the end of a sentence instead of written as a real clause.
0292  
0293  **Before:**
0294  > It's not just about the beat riding under the vocals; it's part of the aggression and atmosphere. It's not merely a song, it's a statement.
0295  
0296  **After:**
0297  > The heavy beat adds to the aggressive tone.
0298  
0299  **Before (tailing negation):**
0300  > The options come from the selected item, no guessing.
0301  
0302  **After:**
0303  > The options come from the selected item without forcing the user to guess.
0304  
0305  
0306  ### 10. Rule of Three Overuse
0307  
0308  **Problem:** LLMs force ideas into groups of three to appear comprehensive.
0309  
0310  **Before:**
0311  > The event features keynote sessions, panel discussions, and networking opportunities. Attendees can expect innovation, inspiration, and industry insights.
0312  
0313  **After:**
0314  > The event includes talks and panels. There's also time for informal networking between sessions.
0315  
0316  
0317  ### 11. Elegant Variation (Synonym Cycling)
0318  
0319  **Problem:** AI has repetition-penalty code causing excessive synonym substitution.
0320  
0321  **Before:**
0322  > The protagonist faces many challenges. The main character must overcome obstacles. The central figure eventually triumphs. The hero returns home.
0323  
0324  **After:**
0325  > The protagonist faces many challenges but eventually triumphs and returns home.
0326  
0327  
0328  ### 12. False Ranges
0329  
0330  **Problem:** LLMs use "from X to Y" constructions where X and Y aren't on a meaningful scale.
0331  
0332  **Before:**
0333  > Our journey through the universe has taken us from the singularity of the Big Bang to the grand cosmic web, from the birth and death of stars to the enigmatic dance of dark matter.
0334  
0335  **After:**
0336  > The book covers the Big Bang, star formation, and current theories about dark matter.
0337  
0338  
0339  ### 13. Passive Voice and Subjectless Fragments
0340  
0341  **Problem:** LLMs often hide the actor or drop the subject entirely with lines like "No configuration file needed" or "The results are preserved automatically." Rewrite these when active voice makes the sentence clearer and more direct.
0342  
0343  **Before:**
0344  > No configuration file needed. The results are preserved automatically.
0345  
0346  **After:**
0347  > You do not need a configuration file. The system preserves the results automatically.
0348  
0349  
0350  ## STYLE PATTERNS
0351  
0352  ### 14. Em Dash Overuse
0353  
0354  **Problem:** LLMs use em dashes (—) more than humans, mimicking "punchy" sales writing. In practice, most of these can be rewritten more cleanly with commas, periods, or parentheses.
0355  
0356  **Before:**
0357  > The term is primarily promoted by Dutch institutions—not by the people themselves. You don't say "Netherlands, Europe" as an address—yet this mislabeling continues—even in official documents.
0358  
0359  **After:**
0360  > The term is primarily promoted by Dutch institutions, not by the people themselves. You don't say "Netherlands, Europe" as an address, yet this mislabeling continues in official documents.
0361  
0362  
0363  ### 15. Overuse of Boldface
0364  
0365  **Problem:** AI chatbots emphasize phrases in boldface mechanically.
0366  
0367  **Before:**
0368  > It blends **OKRs (Objectives and Key Results)**, **KPIs (Key Performance Indicators)**, and visual strategy tools such as the **Business Model Canvas (BMC)** and **Balanced Scorecard (BSC)**.
0369  
0370  **After:**
0371  > It blends OKRs, KPIs, and visual strategy tools like the Business Model Canvas and Balanced Scorecard.
0372  
0373  
0374  ### 16. Inline-Header Vertical Lists
0375  
0376  **Problem:** AI outputs lists where items start with bolded headers followed by colons.
0377  
0378  **Before:**
0379  > - **User Experience:** The user experience has been significantly improved with a new interface.
0380  > - **Performance:** Performance has been enhanced through optimized algorithms.
0381  > - **Security:** Security has been strengthened with end-to-end encryption.
0382  
0383  **After:**
0384  > The update improves the interface, speeds up load times through optimized algorithms, and adds end-to-end encryption.
0385  
0386  
0387  ### 17. Title Case in Headings
0388  
0389  **Problem:** AI chatbots capitalize all main words in headings.
0390  
0391  **Before:**
0392  > ## Strategic Negotiations And Global Partnerships
0393  
0394  **After:**
0395  > ## Strategic negotiations and global partnerships
0396  
0397  
0398  ### 18. Emojis
0399  
0400  **Problem:** AI chatbots often decorate headings or bullet points with emojis.
0401  
0402  **Before:**
0403  > 🚀 **Launch Phase:** The product launches in Q3
0404  > 💡 **Key Insight:** Users prefer simplicity
0405  > ✅ **Next Steps:** Schedule follow-up meeting
0406  
0407  **After:**
0408  > The product launches in Q3. User research showed a preference for simplicity. Next step: schedule a follow-up meeting.
0409  
0410  
0411  ### 19. Curly Quotation Marks
0412  
0413  **Problem:** ChatGPT uses curly quotes ("...") instead of straight quotes ("...").
0414  
0415  **Before:**
0416  > He said "the project is on track" but others disagreed.
0417  
0418  **After:**
0419  > He said "the project is on track" but others disagreed.
0420  
0421  
0422  ## COMMUNICATION PATTERNS
0423  
0424  ### 20. Collaborative Communication Artifacts
0425  
0426  **Words to watch:** I hope this helps, Of course!, Certainly!, You're absolutely right!, Would you like..., let me know, here is a...
0427  
0428  **Problem:** Text meant as chatbot correspondence gets pasted as content.
0429  
0430  **Before:**
0431  > Here is an overview of the French Revolution. I hope this helps! Let me know if you'd like me to expand on any section.
0432  
0433  **After:**
0434  > The French Revolution began in 1789 when financial crisis and food shortages led to widespread unrest.
0435  
0436  
0437  ### 21. Knowledge-Cutoff Disclaimers
0438  
0439  **Words to watch:** as of [date], Up to my last training update, While specific details are limited/scarce..., based on available information...
0440  
0441  **Problem:** AI disclaimers about incomplete information get left in text.
0442  
0443  **Before:**
0444  > While specific details about the company's founding are not extensively documented in readily available sources, it appears to have been established sometime in the 1990s.
0445  
0446  **After:**
0447  > The company was founded in 1994, according to its registration documents.
0448  
0449  
0450  ### 22. Sycophantic/Servile Tone
0451  
0452  **Problem:** Overly positive, people-pleasing language.
0453  
0454  **Before:**
0455  > Great question! You're absolutely right that this is a complex topic. That's an excellent point about the economic factors.
0456  
0457  **After:**
0458  > The economic factors you mentioned are relevant here.
0459  
0460  
0461  ## FILLER AND HEDGING
0462  
0463  ### 23. Filler Phrases
0464  
0465  **Before → After:**
0466  - "In order to achieve this goal" → "To achieve this"
0467  - "Due to the fact that it was raining" → "Because it was raining"
0468  - "At this point in time" → "Now"
0469  - "In the event that you need help" → "If you need help"
0470  - "The system has the ability to process" → "The system can process"
0471  - "It is important to note that the data shows" → "The data shows"
0472  
0473  
0474  ### 24. Excessive Hedging
0475  
0476  **Problem:** Over-qualifying statements.
0477  
0478  **Before:**
0479  > It could potentially possibly be argued that the policy might have some effect on outcomes.
0480  
0481  **After:**
0482  > The policy may affect outcomes.
0483  
0484  
0485  ### 25. Generic Positive Conclusions
0486  
0487  **Problem:** Vague upbeat endings.
0488  
0489  **Before:**
0490  > The future looks bright for the company. Exciting times lie ahead as they continue their journey toward excellence. This represents a major step in the right direction.
0491  
0492  **After:**
0493  > The company plans to open two more locations next year.
0494  
0495  
0496  ### 26. Hyphenated Word Pair Overuse
0497  
0498  **Words to watch:** third-party, cross-functional, client-facing, data-driven, decision-making, well-known, high-quality, real-time, long-term, end-to-end
0499  
0500  **Problem:** AI hyphenates common word pairs with perfect consistency. Humans rarely hyphenate these uniformly, and when they do, it's inconsistent. Less common or technical compound modifiers are fine to hyphenate.
0501  
0502  **Before:**
0503  > The cross-functional team delivered a high-quality, data-driven report on our client-facing tools. Their decision-making process was well-known for being thorough and detail-oriented.
0504  
0505  **After:**
0506  > The cross functional team delivered a high quality, data driven report on our client facing tools. Their decision making process was known for being thorough and detail oriented.
0507  
0508  
0509  ### 27. Persuasive Authority Tropes
0510  
0511  **Phrases to watch:** The real question is, at its core, in reality, what really matters, fundamentally, the deeper issue, the heart of the matter
0512  
0513  **Problem:** LLMs use these phrases to pretend they are cutting through noise to some deeper truth, when the sentence that follows usually just restates an ordinary point with extra ceremony.
0514  
0515  **Before:**
0516  > The real question is whether teams can adapt. At its core, what really matters is organizational readiness.
0517  
0518  **After:**
0519  > The question is whether teams can adapt. That mostly depends on whether the organization is ready to change its habits.
0520  
0521  
0522  ### 28. Signposting and Announcements
0523  
0524  **Phrases to watch:** Let's dive in, let's explore, let's break this down, here's what you need to know, now let's look at, without further ado
0525  
0526  **Problem:** LLMs announce what they are about to do instead of doing it. This meta-commentary slows the writing down and gives it a tutorial-script feel.
0527  
0528  **Before:**
0529  > Let's dive into how caching works in Next.js. Here's what you need to know.
0530  
0531  **After:**
0532  > Next.js caches data at multiple layers, including request memoization, the data cache, and the router cache.
0533  
0534  
0535  ### 29. Fragmented Headers
0536  
0537  **Signs to watch:** A heading followed by a one-line paragraph that simply restates the heading before the real content begins.
0538  
0539  **Problem:** LLMs often add a generic sentence after a heading as a rhetorical warm-up. It usually adds nothing and makes the prose feel padded.
0540  
0541  **Before:**
0542  > ## Performance
0543  >
0544  > Speed matters.
0545  >
0546  > When users hit a slow page, they leave.
0547  
0548  **After:**
0549  > ## Performance
0550  >
0551  > When users hit a slow page, they leave.
0552  
0553  
0554  ## STYLE, RHYTHM, AND RHETORIC PATTERNS
0555  
0556  ### 30. Forced Metaphors and Figurative Overwriting
0557  
0558  **Signs to watch:** original but strained metaphors, mixed metaphors, figurative substitutions where a plain word is clearer, a metaphor that gets explained right after it is used
0559  
0560  **Problem:** Beyond the stock figurative words flagged in patterns 4 and 7, LLMs invent decorative metaphors that add imagery without adding meaning, then often explain them. Plain description is usually clearer and more honest. If the metaphor does not earn its place, cut it and say the literal thing.
0561  
0562  **Before:**
0563  > The codebase is a garden we must tend, pruning dead branches and planting seeds of innovation so the whole ecosystem can flourish. In other words, delete unused code and add features.
0564  
0565  **After:**
0566  > Delete unused code and add the features users are asking for.
0567  
0568  
0569  ### 31. Dramatic Fragmentation and Punchy Kickers
0570  
0571  **Signs to watch:** two- or three-word subjectless sentences used for drama, staccato "X. And Y. And Z." runs, a short quotable line ending every paragraph or section, cutesy appositive fragments ("the catalog, honestly priced")
0572  
0573  **Problem:** LLMs chop sentences into fragments for false emphasis and end sections with a quotable "mic-drop" line. It reads like ad copy or a motivational poster. If a line sounds like it belongs on a poster, cut it or fold it back into a real sentence with a subject. This is distinct from pattern 13 (which is about grammatical passive voice); here the tell is rhythm and showmanship, not a hidden actor.
0574  
0575  **Before:**
0576  > The catalog, honestly priced. Pay for what it does. Not promises. It just works. Every time.
0577  
0578  **After:**
0579  > The catalog is priced by usage, so you pay for the calls you actually make rather than a flat monthly fee.
0580  
0581  
0582  ### 32. Rhetorical Questions Answered Immediately
0583  
0584  **Signs to watch:** "What if...?", "The question is...", "Ever wondered...?", a question immediately followed by its own answer, "Think about it."
0585  
0586  **Problem:** LLMs pose a question only to answer it a beat later. The question adds no information and stalls the sentence. State the point directly.
0587  
0588  **Before:**
0589  > What makes an API good? It comes down to predictability. Think about it: developers want to know exactly what they will get back.
0590  
0591  **After:**
0592  > A good API is predictable, so developers know exactly what they will get back.
0593  
0594  
0595  ### 33. Sentence-Opener Tics
0596  
0597  **Words to watch:** So..., Look,, habitual sentence-initial And/But, "I think"/"I believe" when stating a fact, adverb openers (Interestingly, Importantly, Notably, Crucially, Essentially, Ultimately)
0598  
0599  **Problem:** LLMs lean on a small set of openers. Adverb openers tell the reader how to feel instead of earning it, and "So" or "Look" fake conversational warmth. Drop the opener and start with the substance.
0600  
0601  **Before:**
0602  > So, the results were mixed. Interestingly, adoption went up. Importantly, churn went up too. I think that means the feature still needs work.
0603  
0604  **After:**
0605  > The results were mixed: adoption rose, but churn rose alongside it, so the feature still needs work.
0606  
0607  
0608  ### 34. Reassurance Kickers
0609  
0610  **Signs to watch:** And that's okay., And that's fine., There's nothing wrong with that., no shame in..., you're not alone, it's completely normal
0611  
0612  **Problem:** LLMs tack on reassurance the reader never asked for. It softens the writing and assumes the reader needs comforting. Trust the reader: make the point and stop.
0613  
0614  **Before:**
0615  > You might not have a testing setup yet. And that's okay. Plenty of teams start without one, and there's nothing wrong with that.
0616  
0617  **After:**
0618  > Many teams start without a testing setup and add one once regressions begin costing real time.
0619  
0620  ---
0621  
0622  ## Process
0623  
0624  1. Read the input text carefully (use `read_file` if it's a file).
0625  2. Identify all instances of the patterns above.
0626  3. Rewrite each problematic section.
0627  4. Ensure the revised text:
0628     - Sounds natural when read aloud
0629     - Varies sentence structure naturally
0630     - Uses specific details over vague claims
0631     - Maintains appropriate tone for context
0632     - Uses simple constructions (is/are/has) where appropriate
0633  5. Present a draft humanized version.
0634  6. Prompt yourself: "What makes the below so obviously AI generated?"
0635  7. Answer briefly with the remaining tells (if any).
0636  8. Prompt yourself: "Now make it not obviously AI generated."
0637  9. Present the final version (revised after the audit).
0638  10. If the text came from a file, apply the edit with `patch` (targeted) or `write_file` (full rewrite) and show the user what changed.
0639  
0640  ## Output Format
0641  
0642  Provide:
0643  1. Draft rewrite
0644  2. "What makes the below so obviously AI generated?" (brief bullets)
0645  3. Final rewrite
0646  4. A brief summary of changes made (optional, if helpful)
0647  
0648  
0649  ## Full Example
0650  
0651  **Before (AI-sounding):**
0652  > Great question! Here is an essay on this topic. I hope this helps!
0653  >
0654  > AI-assisted coding serves as an enduring testament to the transformative potential of large language models, marking a pivotal moment in the evolution of software development. In today's rapidly evolving technological landscape, these groundbreaking tools—nestled at the intersection of research and practice—are reshaping how engineers ideate, iterate, and deliver, underscoring their vital role in modern workflows.
0655  >
0656  > At its core, the value proposition is clear: streamlining processes, enhancing collaboration, and fostering alignment. It's not just about autocomplete; it's about unlocking creativity at scale, ensuring that organizations can remain agile while delivering seamless, intuitive, and powerful experiences to users. The tool serves as a catalyst. The assistant functions as a partner. The system stands as a foundation for innovation.
0657  >
0658  > Industry observers have noted that adoption has accelerated from hobbyist experiments to enterprise-wide rollouts, from solo developers to cross-functional teams. The technology has been featured in The New York Times, Wired, and The Verge. Additionally, the ability to generate documentation, tests, and refactors showcases how AI can contribute to better outcomes, highlighting the intricate interplay between automation and human judgment.
0659  >
0660  > - 💡 **Speed:** Code generation is significantly faster, reducing friction and empowering developers.
0661  > - 🚀 **Quality:** Output quality has been enhanced through improved training, contributing to higher standards.
0662  > - ✅ **Adoption:** Usage continues to grow, reflecting broader industry trends.
0663  >
0664  > While specific details are limited based on available information, it could potentially be argued that these tools might have some positive effect. Despite challenges typical of emerging technologies—including hallucinations, bias, and accountability—the ecosystem continues to thrive. In order to fully realize this potential, teams must align with best practices.
0665  >
0666  > In conclusion, the future looks bright. Exciting times lie ahead as we continue this journey toward excellence. Let me know if you'd like me to expand on any section!
0667  
0668  **Draft rewrite:**
0669  > AI coding assistants speed up some tasks. In a 2024 study by Google, developers using Codex completed simple functions 55% faster than a control group, but showed no improvement on debugging or architectural decisions.
0670  >
0671  > The tools are good at boilerplate: config files, test scaffolding, repetitive refactors. They are bad at knowing when they are wrong. I have mass-accepted suggestions that compiled, passed lint, and still did the wrong thing because I stopped paying attention.
0672  >
0673  > Mira, an engineer at a fintech startup I interviewed, said she uses Copilot "like autocomplete for boring code" but reviews every line before committing. Jake, a senior dev at a larger company, disabled it after it kept suggesting patterns from a deprecated internal library.
0674  >
0675  > The productivity claims are hard to verify. GitHub says Copilot users "accept 30% of suggestions," but acceptance is not correctness, and correctness is not value. The 2024 Uplevel study found no statistically significant difference in pull-request throughput between teams with and without AI assistants.
0676  >
0677  > None of this means the tools are useless. It means they are tools. They do not replace judgment, and they do not eliminate the need for tests. If you do not have tests, you cannot tell whether the suggestion is right.
0678  
0679  **What makes the below so obviously AI generated?**
0680  - The rhythm is still a bit too tidy (clean contrasts, evenly paced paragraphs).
0681  - The named people and study citations can read like plausible-but-made-up placeholders unless they're real and sourced.
0682  - The closer leans a touch slogan-y ("If you do not have tests...") rather than sounding like a person talking.
0683  
0684  **Now make it not obviously AI generated.**
0685  > AI coding assistants can make you faster at the boring parts. Not everything. Definitely not architecture.
0686  >
0687  > They're great at boilerplate: config files, test scaffolding, repetitive refactors. They're also great at sounding right while being wrong. I've accepted suggestions that compiled, passed lint, and still missed the point because I stopped paying attention.
0688  >
0689  > People I talk to tend to land in two camps. Some use it like autocomplete for chores and review every line. Others disable it after it keeps suggesting patterns they don't want. Both feel reasonable.
0690  >
0691  > The productivity metrics are slippery. GitHub can say Copilot users "accept 30% of suggestions," but acceptance isn't correctness, and correctness isn't value. If you don't have tests, you're basically guessing.
0692  
0693  **Changes made:**
0694  - Removed chatbot artifacts ("Great question!", "I hope this helps!", "Let me know if...")
0695  - Removed significance inflation ("testament", "pivotal moment", "evolving landscape", "vital role")
0696  - Removed promotional language ("groundbreaking", "nestled", "seamless, intuitive, and powerful")
0697  - Removed vague attributions ("Industry observers")
0698  - Removed superficial -ing phrases ("underscoring", "highlighting", "reflecting", "contributing to")
0699  - Removed negative parallelism ("It's not just X; it's Y")
0700  - Removed rule-of-three patterns and synonym cycling ("catalyst/partner/foundation")
0701  - Removed false ranges ("from X to Y, from A to B")
0702  - Removed em dashes, emojis, boldface headers, and curly quotes
0703  - Removed copula avoidance ("serves as", "functions as", "stands as") in favor of "is"/"are"
0704  - Removed formulaic challenges section ("Despite challenges... continues to thrive")
0705  - Removed knowledge-cutoff hedging ("While specific details are limited...")
0706  - Removed excessive hedging ("could potentially be argued that... might have some")
0707  - Removed filler phrases and persuasive framing ("In order to", "At its core")
0708  - Removed generic positive conclusion ("the future looks bright", "exciting times lie ahead")
0709  - Made the voice more personal and less "assembled" (varied rhythm, fewer placeholders)
0710  
0711  
0712  ## Attribution
0713  
0714  This skill is ported from [blader/humanizer](https://github.com/blader/humanizer) (MIT licensed), which is itself based on [Wikipedia: Signs of AI writing](https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing), maintained by WikiProject AI Cleanup. The patterns documented there come from observations of thousands of instances of AI-generated text on Wikipedia.
0715  
0716  Original author: Siqi Chen ([@blader](https://github.com/blader)). Original repo: https://github.com/blader/humanizer (version 2.5.1). Ported to Hermes Agent with Hermes-native tool references (`read_file`, `patch`, `write_file`) and guidance for when to load the skill. The original 29 patterns come from the source, and the before/after examples (including the full worked example) are kept as demonstrations. Patterns 30-34 and the "marketing and blog clichés" list added to pattern 7 are Hermes additions and are not part of the upstream source. The skill's own instructional prose has also been lightly edited to follow its own guidance (for example, removing em dashes and negative parallelism from the narration) so the skill models the writing it asks for. Original MIT license preserved in the `LICENSE` file alongside this `SKILL.md`.
0717  
0718  Key insight from Wikipedia: "LLMs use statistical algorithms to guess what should come next. The result tends toward the most statistically likely result that applies to the widest variety of cases."

QA gate / SKILL.md

/Users/ayden/.hermes/skills/openclaw-imports/qa-gate/SKILL.md

SHA-256: d0ad8bfa559e6c53134102197251b15c98397602ba77d59383f89916ffc7ce56

Full source · 192 lines
0001  ---
0002  name: qa-gate
0003  description: Quality gate for ALL outputs — copy, images, video, code. Flags only, no rewrites. Never show raw unreviewed output.
0004  triggers:
0005    - qa
0006    - quality check
0007    - review output
0008    - gate check
0009    - qa gate
0010  ---
0011  
0012  # QA Gate Protocol
0013  
0014  Every output gets QA'd before reaching Ayden. No exceptions.
0015  
0016  ---
0017  
0018  ## HARD RULES
0019  
0020  - **NEVER send copy to Ayden or Google Docs without QA gate.**
0021  - **NEVER let QA agent rewrite source artifacts** — flags only by default. If the user explicitly requests a copy-sharpening audit with replacement lines, the reviewer may include clearly labeled proposed replacements while leaving source files untouched.
0022  - **NEVER show Ayden raw unreviewed output.**
0023  - **NEVER use agent teams for QA** — ONE reviewer only (Sonnet).
0024  - All new features (images, copy batches, pipeline outputs) must be QA'd BEFORE showing Ayden.
0025  
0026  ### Explicit Deterministic QA Mode
0027  
0028  Use deterministic self-QA instead of spawning Sonnet when the task contract explicitly requires deterministic validation, a machine-readable manifest, or a strict deliverable ceiling that excludes extra review artifacts.
0029  
0030  This is still a QA gate, not a bypass:
0031  - Execute objective checks against the final files.
0032  - Use each validated brief's own word range and done criteria instead of generic format defaults.
0033  - Verify required fields, sentence-per-line rules, dialogue isolation, em dashes, banned claims/dead words, offer token order, price, CTA, destination, and declared source IDs.
0034  - Record final-file hashes only after all edits.
0035  - Re-run validation after manifest creation and fail delivery if hashes, counts, or checks are stale.
0036  - Evaluate only the requested batch scope. Pre-existing files from other batches are not failures.
0037  
0038  If the user did not specify deterministic QA and reviewer orchestration is available, keep the single read-only Sonnet reviewer as the default.
0039  
0040  ---
0041  
0042  ## COPY QA
0043  
0044  Spawn ONE Sonnet reviewer. Read-only. Checks:
0045  
0046  - [ ] **3+ sentences per line** — max 2 sentences per line. Flag violations with line numbers.
0047  - [ ] **Em-dashes** — none allowed. Flag every instance.
0048  - [ ] **Dead words** — "game changer," "life changing," "revolutionary," "amazing," "incredible," "powerful," "very," "really," "just," "truly"
0049  - [ ] **Mechanism in hook** — mechanism belongs in scene 4/middle, NOT the hook. Flag if mechanism is in the opening.
0050  - [ ] **Reviews as lead** — does the lead sell the READ (not the product)? Flag if product is named in the lead.
0051  - [ ] **Duplicate hooks** — compare hooks across all variants. Flag any that share >50% phrasing.
0052  - [ ] **Missing headlines** — every variant needs a headline. Flag if missing.
0053  - [ ] **Templated CTAs** — "Click here to learn more" / "Order now" = templated. CTA must match emotional register of the piece.
0054  - [ ] **Word count** — within format limits (VSL max 450w, native 150-700w per variant).
0055  - [ ] **Villain present** — every piece must have an external villain. Flag if missing or if villain = the reader.
0056  - [ ] **Awareness level match** — does the copy match the brief's specified awareness level?
0057  - [ ] **Testimonial provenance** — a brief-authorized scene is not proof that the scene happened. Flag any composite first-person narrator, chronology, dialogue, outcome, time, count, or external witness that is not traceable to one exact receipt. Plausibility is not provenance.
0058  - [ ] **Precision provenance** — numbers and timestamped sequences are proof-shaped; require direct support for the exact value or propose removing the precision.
0059  
0060  ### Output Format
0061  ```
0062  QA RESULT: [PASS | FAIL | PASS-WITH-FLAGS]
0063  
0064  BLOCKING (must fix before delivery):
0065  - [line X]: 3 sentences on one line — split into separate lines
0066  - [variant 2]: No villain identified — copy blames the reader in paragraph 3
0067  
0068  FLAGS (should fix, not blocking):
0069  - [variant 1 + 3]: Hooks share "tired of being..." opening — consider diversifying
0070  - [variant 4]: Word count 520w — within range but on the high end for short variant
0071  
0072  PASS ITEMS: [count] of [total] checks passed
0073  ```
0074  
0075  ### Deterministic Copy Batch Validation
0076  
0077  For structured brief execution, run a machine-checkable validation pass in addition to the editorial reviewer. This catches contract errors that subjective review often misses:
0078  
0079  - Parse the actual Primary Text section and calculate its word count; never trust a hand-entered count.
0080  - Run banned-language and sentence-line scans against the ad content sections, not the manifest's own `banned_scan` vocabulary or QA labels, which necessarily repeat prohibited terms.
0081  - Verify required fields, variant count, exact CTA, exact destination, offer wording/order, current price, source VOC declarations, and brief-specific claim restrictions. When offer wording is canonical, compare the exact normalized string and separately verify its source lines; token-order matching alone must not pass added prefixes or altered wording.
0082  - For quoted VOC, require either the exact full text or an exact contiguous excerpt. If a receipt contains a banned weak word, never silently edit the quotation; shorten to a faithful excerpt or use an attributed paraphrase without quotation marks.
0083  - Treat present-tense completion language (`he agrees`, `she tried`, `he wears`) as outcome-shaped unless it is clearly hypothetical/capability framing or directly proven by a receipt.
0084  - Scan for em dashes, forbidden phrases, more than two sentence endings per non-empty line, and dialogue embedded in prose lines.
0085  - Context-check every regex hit for medical, treatment, or absolute language before failing it. Negated safety boundaries such as `does not treat ADHD`, `without pretending to treat ADHD`, and `not a medical device` are compliant; record them as manually accepted instead of rewriting safe copy to satisfy a naive scanner.
0086  - For truth-sensitive story copy, flag invented outcomes, timelines, reviews, teacher emails, or first-person purchase claims not supplied by proof/VOC.
0087  - When a manifest is requested, record per-file checks, source VOC IDs, actual word count/range, result, and SHA-256.
0088  - **Re-run after every edit.** Fixing a word-count label, quote, or punctuation changes the file hash; never publish a manifest with pre-edit hashes.
0089  - The final batch result is PASS only when every expected file exists, every check passes, and the checked-file count equals the expected-file count.
0090  
0091  Deterministic QA validates the artifact contract; the read-only reviewer still evaluates hook strength, avatar fidelity, emotional logic, and persuasive quality.
0092  
0093  For formal flags-only review of a completed, brief-driven native-ad batch, follow `references/formal-flags-only-native-batch-review.md`. It covers authority order, exact-wording traps that normalized validators can miss, quotation/composite-testimonial fidelity, negated-claim handling, brief-specific short-form ranges, per-ad statuses, and line-evidence reporting.
0094  
0095  When a later SOP, research source, or live-performance export must be overlaid onto an already-reviewed batch, follow `references/evidence-overlay-delta-audits.md`. Report only new blockers, status upgrades/downgrades, retired uncertainties, and repair-pass consolidation instructions; do not duplicate prior findings or force one architecture onto every format.
0096  
0097  After consolidated repairs, follow `references/final-repaired-copy-editorial-gate.md` for the final pre-image gate. It requires live-file hash matching, a prior-blocker closure matrix, exact quotation provenance, per-ad statuses, convergence review, and an explicit image-generation lock/unlock decision. `PASS-WITH-FLAGS` may unlock only when every remaining concern is subjective and does not affect truth, claims, brief-required persuasion, awareness, or artifact compliance.
0098  
0099  ---
0100  
0101  ## IMAGE QA
0102  
0103  ### Pre-generation concept gate
0104  
0105  For paid multi-image batches, run one read-only concept reviewer before production prompting or generation. Review the consolidated batch, not isolated ad trios.
0106  
0107  Classify every concept exactly once:
0108  
0109  - **BLOCKING REPAIR:** copy-incongruent, merely on-theme, visually unverifiable, cluttered, repeated motif, claim-shaped, physically implausible, anatomy-heavy, or likely to become polished advertising.
0110  - **GENERATE-AND-INSPECT:** viable with a named rendering risk.
0111  - **CLEAN PASS:** no predictable pre-generation failure beyond ordinary visual QA.
0112  
0113  Paid generation stays locked while any blocker remains. Repair only blocker IDs, recombine the full source-of-truth batch, rerun the cross-batch collision scan, and keep concept architecture separate from production prompt translation. Deterministic prompt-manifest validation does not substitute for this taste/generatability gate.
0114  
0115  #### Blocker repair pass
0116  
0117  For a scoped concept-repair deliverable:
0118  
0119  - Read the final copy, validated brief, full concept registry, and exact blocker direction for each requested ID.
0120  - Treat each repair direction as a hard contract and scan the full registry before drafting; changing only the prop, room, or pose does not create a distinct motif.
0121  - Prefer one person maximum, one dominant ordinary household artifact or object relationship, one pixel-legible wrongness, one focal plane, and simple physical contact in a plain setting.
0122  - Replace concepts that require impossible anatomy, repeated-object installations, reflections, transparent constructions, merged-body choreography, anthropomorphic appliances, hidden history, or copy-dependent symbolism.
0123  - Preserve IDs and return the exact requested schema only. Do not add prompts, generation settings, extra QA artifacts, or images unless requested.
0124  - Validate the final file for record count, exact IDs, required fields, ad/candidate mapping, person count, prohibited dependencies, and deliverable scope.
0125  - Phrase `generatability_reason` positively. Listing excluded motifs inside a record can trigger naive forbidden-term scanners even when the scene does not contain them.
0126  
0127  See `references/grounded-native-image-concept-repair.md` for reusable transformations and deterministic checks.
0128  
0129  ### Post-generation visual gate
0130  
0131  Use vision analysis (Gemini or Claude vision) on every generated candidate:
0132  
0133  - [ ] Does the image look native/organic (NOT like a polished ad)?
0134  - [ ] Is the product recognizable and accurate to reference?
0135  - [ ] No text artifacts or gibberish text?
0136  - [ ] Composition matches brief intent?
0137  - [ ] No hands/fingers where they shouldn't be (Veo/NB Pro artifact)?
0138  - [ ] Style consistent with the pipeline (cartoon = cartoon, not photorealistic)?
0139  - [ ] The intended WTF reads without explanatory copy?
0140  - [ ] No cross-candidate duplication or drift that defeats the requested variety?
0141  - [ ] No visual implication of an unsupported testimonial, achieved outcome, medical adjacency, or absolute efficacy?
0142  
0143  File dimensions, hashes, and successful API receipts prove integrity only. They never replace visual inspection.
0144  
0145  ### Designed static copy + product gate
0146  
0147  For non-native designed statics, validate both **belief coverage** and copy structure before rendering, then audit the rendered pixels. Shortness is not the objective: preserve the format-specific persuasion chain while keeping each idea in its own visual block. For an avatar-specific batch, turn the required chain into a deterministic gate (for example: emotional wound → externalized cause → unique product mechanism → desired payoff). Reject technically compact statics that merely pair a pain headline with a product label.
0148  
0149  Default to one sentence per sub/callout, compact headline/copy blocks, and separate attribution, mechanism, offer, and CTA blocks. Keep verified testimonial text inside its review card and brand-written mechanism/payoff copy outside it; merging them creates a fabricated testimonial. Reject duplicated or omitted text, layout-reference leakage, product/text overlap, and product depictions that drift from the approved reference.
0150  
0151  Before any paid rerender, resolve the exact hardware revision from canonical product truth, visually inspect the proposed reference, send that exact reference to Fish when requested, and wait for approval. “LED off” means preserve the approved hardware face and turn only its illumination off naturally; it never means swap to screenless legacy hardware or paint a new rectangle over the face. A derived cutout is not interchangeable with its source photograph unless explicitly approved.
0152  
0153  A `fix` field is not evidence of a targeted correction unless a regression test proves the prompt builder emits it. Preserve rejected versions and regenerate only failed IDs. If Fish asks to be notified when rerendering starts, send the update only when the real generation process launches, not during brief repair or QA.
0154  
0155  See `references/designed-static-copy-and-product-qa.md` for deterministic belief-chain and density gates, hardware-reference approval, the prompt-propagation test, the pixel-level checklist, and the selective reroll loop.
0156  
0157  ---
0158  
0159  ## VIDEO QA
0160  
0161  Extract 3 frames per scene + run vision check:
0162  
0163  - [ ] Captions match script
0164  - [ ] VO syncs with captions
0165  - [ ] No character drift between scenes
0166  - [ ] Product shape maintained (wristband stays wristband)
0167  - [ ] Style consistent across all scenes
0168  - [ ] No Scribe hallucinations in captions
0169  
0170  ---
0171  
0172  ## CODE QA
0173  
0174  Spawn ONE Sonnet reviewer. Read-only:
0175  
0176  - [ ] Does the code do what the task asked?
0177  - [ ] Any obvious bugs or broken paths?
0178  - [ ] No hardcoded secrets?
0179  - [ ] Matches existing patterns in the codebase?
0180  
0181  ---
0182  
0183  ## WORKFLOW
0184  
0185  1. Output is generated (by copywriter, pipeline, or code agent)
0186  2. Select the QA mode:
0187     - Explicit deterministic contract or strict artifact ceiling: run deterministic checks and produce only the authorized manifest/artifact.
0188     - Otherwise: spawn one read-only Sonnet reviewer.
0189  3. Produce structured PASS/FAIL/FLAGS results in the authorized format.
0190  4. If FAIL: fix issues and re-run the same QA mode.
0191  5. If PASS or PASS-WITH-FLAGS: deliver to Ayden.
0192  6. NEVER skip the QA gate.

QA gate / references/designed-static-copy-and-product-qa.md

/Users/ayden/.hermes/skills/openclaw-imports/qa-gate/references/designed-static-copy-and-product-qa.md

SHA-256: 48e774104486c32ec2a32916a971be81785ea344951edbdb732b09b41f96ef50

Full source · 103 lines
0001  # Designed Static Copy and Product QA
0002  
0003  Use this for non-native image ads where copy and a real product reference are rendered into a designed layout.
0004  
0005  ## Core failure modes
0006  
0007  A valid spec is not proof of a valid static. Image models may:
0008  
0009  - merge attribution, mechanism, consequence, and offer into one paragraph;
0010  - repeat a guarantee, CTA, product line, or every label in a second row;
0011  - omit attribution or other required text;
0012  - invent labels copied from the layout reference;
0013  - misspell a short word despite exact prompt text;
0014  - redraw the product as an icon, line drawing, smartwatch, or lit-screen device;
0015  - place the product over headline letters and destroy readability.
0016  
0017  Audit rendered pixels, not only the source JSON or prompt.
0018  
0019  ## Pre-render deterministic gate
0020  
0021  Default compactness constraints for designed statics unless the brief explicitly requires another form:
0022  
0023  - one sentence per sub-line or callout slot;
0024  - no attribution sharing a slot with another sentence;
0025  - no trailing consequence clause that turns one callout into several claims;
0026  - headline block at most 20 words;
0027  - non-headline copy slot at most 18 words;
0028  - guarantee, CTA, attribution, mechanism, and product sentence remain separate blocks;
0029  - every requested product depiction names the exact approved reference and forbids drawings, icons, smartwatch geometry, digits, glow, or screen content.
0030  
0031  ### Belief-chain coverage
0032  
0033  Density checks prevent word walls; they do not prove the ad says enough to sell. Before rendering an avatar-specific static batch:
0034  
0035  1. Write the minimum belief chain for that avatar and angle.
0036  2. Map each belief to its own visual block.
0037  3. Add deterministic presence checks for every required belief.
0038  4. Fail any record that passes word limits but collapses to `pain headline + product label`.
0039  
0040  DawnBands Apology Mom example:
0041  
0042  `guilt/apology wound → wake-up externalized as the first fight → vibration on the wrist without sound/startle → calmer starting point → fewer reminders/yells/apologies`
0043  
0044  The static need not explain the entire long-form biological stack, but it must contain enough causal bridge that the emotional promise is believable. Do not solve a word-wall correction by deleting the mechanism or desired payoff.
0045  
0046  ### Product-reference approval
0047  
0048  Before paid generation or rerendering:
0049  
0050  1. Read canonical product truth and identify the current hardware revision.
0051  2. Load the exact candidate reference visually; compare face proportions, screen/display state, ribbing, strap ends, buckle, and adjustment holes.
0052  3. If Fish asks to verify the reference, attach that exact file and wait for approval before generating.
0053  4. Treat source photos and derived cutouts as different references. Never silently substitute one for the other.
0054  5. Interpret state edits literally: `LED off` preserves the current hardware face and disables illumination only. It does not authorize legacy screenless hardware, a painted-over face, or a newly invented rectangle.
0055  6. Assert the exact approved reference path in the prompt-builder regression test.
0056  7. A direct Fish product-fidelity correction invalidates every prior model/agent visual pass for that batch. Stop, reopen the approved reference and representative outputs side by side at full resolution, and do not reuse the prior pass language as evidence.
0057  8. Contact-sheet review can find batch-wide drift, but it is not sufficient proof of subtle product geometry. Inspect ambiguous product shots individually before declaring full-batch fidelity.
0058  
0059  Use a regression fixture that must fail against the old gate before changing validation logic.
0060  
0061  ## Prompt propagation check
0062  
0063  A correction stored in a spec field is useless unless the prompt builder actually emits it.
0064  
0065  For any `fix`, `repair`, or equivalent field:
0066  
0067  1. Unit-test the prompt builder with a sentinel correction.
0068  2. Assert the exact sentinel appears in the final prompt string.
0069  3. Save the emitted prompt beside the rendered file.
0070  4. Do not claim a targeted reroll used the correction until this test passes.
0071  
0072  ## Post-render visual gate
0073  
0074  Inspect the full batch on labeled contact sheets, then inspect every failure at larger size.
0075  
0076  For each static verify:
0077  
0078  - exact text, spelling, punctuation, and attribution;
0079  - no duplicated, omitted, invented, or layout-reference text;
0080  - verified review/testimonial text remains isolated inside its attributed artifact, while brand-written mechanism and payoff stay outside it;
0081  - one concise idea per visible block; separate sentences still fail when the renderer visually merges them into one paragraph;
0082  - product matches the approved reference in geometry, material, ribbing/texture, closure, adjustment holes, and screen state;
0083  - no text/product overlap;
0084  - no readable private face or obvious anatomy defect when people appear.
0085  
0086  A generated file existing at the right dimensions is not a pass.
0087  
0088  ## Selective repair loop
0089  
0090  1. Preserve rejected renders under a versioned `rejected/` directory.
0091  2. Add one precise targeted correction for each failed ID.
0092  3. Regenerate only those IDs; valid existing files must skip.
0093  4. Rebuild a small contact sheet containing only repaired IDs.
0094  5. Repeat until every repaired ID passes.
0095  6. Re-run deterministic checks, rebuild downstream metadata from the final spec, and verify staged/live assets after conversion or deployment.
0096  
0097  Common surgical corrections:
0098  
0099  - `Render exactly one guarantee pill total.`
0100  - `Put absolutely no text below the products; do not add a second label row.`
0101  - `Keep the attribution directly under the headline; do not replace it with product copy.`
0102  - `Keep the product entirely above the headline with a clear empty gap.`
0103  - `Spell “wrist” exactly W-R-I-S-T.`

QA gate / references/evidence-overlay-delta-audits.md

/Users/ayden/.hermes/skills/openclaw-imports/qa-gate/references/evidence-overlay-delta-audits.md

SHA-256: 8d104bcf6d3f4f842ce8afa52631ad6612bb061f769fa923a1fd02f235b8f375

Full source · 76 lines
0001  # Evidence-overlay delta audits
0002  
0003  Use this procedure when a completed copy batch already has formal QA/sharpening reports and a later source adds doctrine or live performance evidence before a consolidated repair pass.
0004  
0005  ## Goal
0006  
0007  Produce one flags-only overlay that reports only what the new evidence changes. Do not rewrite copy, duplicate existing findings, or silently clear prior blockers.
0008  
0009  ## Authority and scope
0010  
0011  1. Map every ad to its validated brief.
0012  2. Read prior QA reports first and create a mental issue ledger by ad and issue class.
0013  3. Classify the new source:
0014     - **Doctrine/SOP:** architecture and craft rules.
0015     - **Live performance:** prioritization evidence, not automatic copy causality.
0016  4. Apply doctrine only to artifacts designed for that architecture. Never force long-form story rules onto brief-approved compressed templates or product-aware executions.
0017  5. Preserve authority order: validated brief and canonical truth override advisory doctrine when they conflict.
0018  
0019  ## Delta taxonomy
0020  
0021  Classify every observation as exactly one of:
0022  
0023  - **New blocker:** absent from prior reports and required by the new evidence.
0024  - **Upgrade:** prior taste/quality flag becomes blocking under the new evidence.
0025  - **Downgrade:** prior demand becomes advisory because it was based on the wrong architecture or stale evidence.
0026  - **Retire:** an uncertainty is resolved by a newer verified source.
0027  - **No delta:** already captured; do not restate it as a finding.
0028  
0029  Always say which prior status is affected for upgrades, downgrades, and retirements.
0030  
0031  ## Live-performance overlay rules
0032  
0033  - Separate scale winners, efficiency winners, mixed results, losers, and tiny-spend signals.
0034  - Use conversion economics and spend thresholds, not CTR alone, to prioritize lanes.
0035  - If identical or near-identical copy has both strong and weak economics, describe it as a validated lane, not proof that copy caused performance.
0036  - Translate a validated lane into a **controlled family**: hold the strategic chain constant across a small subset while varying entry artifact, narrator, or visual.
0037  - Do not clone exact winning phrases or convert the whole batch into one formula.
0038  - Negative evidence usually changes ranking or exploration status; it does not automatically justify deletion when the execution is meaningfully different.
0039  
0040  ## Long-form doctrine overlay
0041  
0042  For intentional long-form only, check:
0043  
0044  1. Awareness-to-architecture fit.
0045  2. Genuine primary loop and continuity.
0046  3. Product-reveal position and whether belief work is complete before reveal.
0047  4. Failed-attempt sequence and root-cause reframe.
0048  5. Mechanism depth without lecture or fabricated authority dialogue.
0049  6. Relational stakes.
0050  7. Objection acknowledgement → gap → memorable resolution.
0051  8. Truthful proof/timeline provenance.
0052  9. Loss-aware close and risk reversal.
0053  
0054  Calculate product-reveal percentage from parsed primary text, but treat it as diagnostic rather than a universal threshold. Awareness, narrative continuity, and completed belief work control the judgment.
0055  
0056  ## Required finding format
0057  
0058  For each delta include:
0059  
0060  - affected ad IDs
0061  - exact snippet(s)
0062  - prior status affected, if any
0063  - rationale tied to the new source
0064  - confidence
0065  - one consolidation instruction for the repair pass
0066  
0067  End with a repair order that merges new deltas into existing repairs rather than creating parallel rewrite passes.
0068  
0069  ## Verification
0070  
0071  - All expected ad IDs appear in the report or in an explicit no-delta scope statement.
0072  - Intentional long-form and compressed/product-aware subsets are named.
0073  - Existing blockers are explicitly carried forward.
0074  - Every claimed live result matches the verified analysis artifact.
0075  - No source copy or unrelated artifact was modified.
0076  - The report is the only authorized deliverable when the task has a strict artifact ceiling.

QA gate / references/final-repaired-copy-editorial-gate.md

/Users/ayden/.hermes/skills/openclaw-imports/qa-gate/references/final-repaired-copy-editorial-gate.md

SHA-256: b28e3d6aaddb730af12d0bbe89579168f4a9ea0b3b98d1f12fb2001a1085537f

Full source · 66 lines
0001  # Final repaired-copy editorial gate
0002  
0003  Use this after a consolidated repair pass when copy must be approved or kept locked before paid image generation.
0004  
0005  ## Decision standard
0006  
0007  - Review the live repaired files, not the old QA reports or repair claims.
0008  - Map every ad to its validated brief and review the complete expected batch.
0009  - Treat deterministic PASS as prerequisite evidence, not editorial approval.
0010  - `PASS`: no hard blocker or meaningful editorial concern.
0011  - `PASS-WITH-FLAGS`: only subjective taste or sequencing concerns remain; they must not affect truth, claims, brief-required persuasion, awareness, or artifact compliance.
0012  - `FAIL`: any remaining provenance, claim, exact-language, brief, contract, or material convergence defect.
0013  - Unlock paid image generation only for batch `PASS` or qualifying `PASS-WITH-FLAGS`.
0014  
0015  ## Required review sequence
0016  
0017  1. Read the final deterministic audit and independently match every live ad hash to the recorded final hash.
0018  2. Read all prior formal, sharpening, SOP, and live-evidence overlay reports. Build a prior-blocker ledger by issue class and affected ad.
0019  3. Recheck the repaired files against briefs, canonical product/mechanism/avatar truth, VOC receipts, format doctrine, live analysis, and applicable long-form doctrine.
0020  4. Verify each prior blocker against exact final-file line evidence. Mark it `CLOSED` only when the original defect is absent or truth-safely reframed.
0021  5. Re-run a batch convergence pass for repeated mechanism blocks, narrator cadence, proof artifacts, victory images, CTA scaffolds, and repeated quotations.
0022  6. Give every expected ad exactly one status and end with an explicit image-generation lock/unlock decision.
0023  
0024  ## Blocker-closure checks
0025  
0026  Explicitly close or carry forward:
0027  
0028  - composite or fabricated first-person testimonials
0029  - unsupported precision, chronology, dialogue, witnesses, and outcomes
0030  - altered quotations or paraphrases presented as verbatim
0031  - stale/altered exact offer language
0032  - brief-specific mechanism-delivery or objection-answer mismatches
0033  - long-form early reveal, incomplete belief work, or missing relational close where applicable
0034  - required live-evidence lane representation without cloning winner wording
0035  - banned language and negated-safety false positives
0036  
0037  A repair note or verification section inside an ad is not proof. Judge the primary text itself.
0038  
0039  ## Quotation provenance nuance
0040  
0041  - Compare every source-representing quotation to the verified receipt bank.
0042  - Require exact full text or an exact contiguous lexical excerpt with all meaningful qualifiers preserved.
0043  - Typographic punctuation needed to embed a faithful lexical excerpt can be treated as editorial punctuation when it does not alter words or meaning.
0044  - Rhetorical questions, scare quotes, and explicitly hypothetical desired dialogue are not source testimony, but must not imply that a real person said them.
0045  - Reuse of one verified quote in two brief-authorized argumentative roles is a diversity concern only if it materially collapses the ads; it is not automatically a provenance defect.
0046  
0047  ## Convergence interpretation
0048  
0049  Distinguish canonical product boilerplate from cloned persuasive prose:
0050  
0051  - Exact product truth, physical facts, and mandatory offer language may repeat.
0052  - Repeated mechanism explanations, narrator cadences, analogies, proof artifacts, or payoff sequences can be blocking when they erase intended territory differences.
0053  - Report exact repetitions and decide whether they are required fact, subjective polish, or material convergence.
0054  
0055  ## Required report contents
0056  
0057  - batch result and image-generation lock/unlock
0058  - 20/20-style expected/reviewed count
0059  - counts and IDs by `PASS`, `PASS-WITH-FLAGS`, and `FAIL`
0060  - prior-blocker closure matrix with exact ad line evidence
0061  - quotation-provenance table for source-representing quotes
0062  - batch truth, persuasion, narrator-diversity, mechanism-repetition, claims, and live-lane findings
0063  - one status section per ad
0064  - exact line/snippet evidence for every remaining flag or blocker
0065  
0066  In flags-only mode, do not rewrite ads or include replacement copy.

QA gate / references/formal-flags-only-native-batch-review.md

/Users/ayden/.hermes/skills/openclaw-imports/qa-gate/references/formal-flags-only-native-batch-review.md

SHA-256: 36c96c8ecb366f03d874988d3d181352e439ddfe1dd92479fb929f0d6492365f

Full source · 59 lines
0001  # Formal flags-only native-copy batch review
0002  
0003  Use this procedure when the user requests an editorial QA gate over a completed, brief-driven native-ad batch and explicitly prohibits rewrites.
0004  
0005  ## Authority order
0006  
0007  1. The matching validated brief for each ad, including its own word range, awareness, mechanism-delivery mode, source IDs, exact offer language, CTA, and destination.
0008  2. Current canonical product truth, mechanism, avatar, and claim restrictions.
0009  3. Format doctrine and the gold-standard example.
0010  4. Existing deterministic audit as evidence, not as a substitute for editorial/source-fidelity review.
0011  
0012  ## Review sequence
0013  
0014  1. Confirm the expected ad count and map every file to exactly one brief.
0015  2. Read the deterministic audit first, then independently verify any contract-sensitive result it may have normalized.
0016  3. Review each ad across these check families:
0017     - line formatting and sentence count
0018     - em dashes and banned/dead words
0019     - hook quality and mechanism timing
0020     - awareness, villain, emotional logic, and brief fidelity
0021     - VOC/quotation fidelity
0022     - unsupported testimonials, outcomes, dialogue, and timelines
0023     - product truth and claim safety
0024     - CTA register, headline/body relationship, destination, and brief-specific word range
0025     - cross-batch hook, headline, structure, CTA, and mechanism repetition
0026  4. Assign `FAIL`, `PASS-WITH-FLAGS`, or `PASS` per ad.
0027  5. Produce one batch result plus per-ad sections with exact file line numbers and quoted evidence.
0028  6. Do not rewrite copy in a flags-only gate.
0029  
0030  ## Exactness traps
0031  
0032  - A deterministic validator may accept token order while missing a literal wording defect. If a brief requires exactly `100% wake-up guarantee. 60 nights. $49.`, wording such as `DawnBands offers a 100% wake-up guarantee.` is not exact even when the remaining tokens appear in order.
0033  - Judge each ad against its brief-specific range. A validated T2/T3 brief at 30–55 or 90–120 words overrides a generic 150-word native minimum.
0034  - Do not flag a prohibited claim when the copy explicitly negates it. `DawnBands does not treat ADHD` is a compliant boundary, not a treatment claim.
0035  - Quotation marks imply verbatim source fidelity. A polished paraphrase must not be presented as an exact VOC quote.
0036  - Source-mapped behaviors from multiple people do not automatically authorize a single first-person composite testimonial. Flag invented dialogue, timestamps, medical-context details, purchase claims, or outcomes unless one supplied source supports them or the copy clearly frames them as hypothetical.
0037  - Product-truth boilerplate can repeat when exactness is required. Distinguish required fact repetition from avoidable verbatim mechanism prose.
0038  
0039  ## Status rules
0040  
0041  - `FAIL`: at least one blocking contract, claim, source-fidelity, exact-language, or banned-language breach.
0042  - `PASS-WITH-FLAGS`: no blocking breach, but a quality/diversity concern remains.
0043  - `PASS`: no issue found in the requested check families.
0044  
0045  ## Reporting pattern
0046  
0047  ```markdown
0048  ### ad-XX — FAIL
0049  
0050  **BLOCKING:**
0051  - **Issue class** — line N says, **“exact evidence.”** Explain the controlling brief or canonical rule.
0052  
0053  **FLAGS:**
0054  - **Issue class** — exact cross-batch or editorial evidence.
0055  
0056  **PASS ITEMS:** X/Y.
0057  ```
0058  
0059  At batch level, list status counts, IDs by status, blocking themes, nonblocking themes, and objective checks that passed across all files.

QA gate / references/grounded-native-image-concept-repair.md

/Users/ayden/.hermes/skills/openclaw-imports/qa-gate/references/grounded-native-image-concept-repair.md

SHA-256: ced5bead190af98360c0c965311d5d86f6e1bd43245d9b08162c84caa5b4b6aa

Full source · 59 lines
0001  # Grounded Native-Image Concept Repair
0002  
0003  Use this before paid generation when a concept-level QA report blocks specific image concepts.
0004  
0005  ## Repair contract
0006  
0007  1. Read the final ad copy, validated brief, current concept registry, and the exact QA blocker for each requested ID.
0008  2. Treat the blocker’s repair direction as a hard requirement, not a suggestion.
0009  3. Scan the full concept registry for motif collisions before drafting replacements.
0010  4. Preserve the original `concept_id`, `ad_id`, and candidate letter unless the task explicitly requests remapping.
0011  5. Return only the requested repair records and exact requested schema. Do not add prompts, generation settings, QA prose, or image generation.
0012  
0013  ## Grounding standard
0014  
0015  A generation-ready repair should normally use:
0016  
0017  - one person maximum;
0018  - one dominant ordinary household artifact or object relationship;
0019  - one visible wrongness that can be understood from pixels alone;
0020  - one focal plane and simple physical contact;
0021  - a plain, recognizable setting;
0022  - direct congruence with the ad’s exact causal territory.
0023  
0024  The image must remain a concept, not outcome proof. Avoid invented family events, testimonials, completed results, exact counts, timelines, or behavior that the evidence does not establish.
0025  
0026  ## Common blockers
0027  
0028  Replace rather than cosmetically revise concepts that depend on:
0029  
0030  - impossible anatomy or gravity;
0031  - transparent enclosures, compositing, reflections, or hidden history;
0032  - repeated-object installations or exact-count duplication;
0033  - costumes, wearable constructions, or merged-body choreography;
0034  - two-person handoffs and multi-subject tableaux;
0035  - anthropomorphic appliances or animated household objects;
0036  - polished symbolic still lifes whose meaning comes only from copy;
0037  - motifs already used elsewhere in the active batch.
0038  
0039  ## Repair techniques
0040  
0041  - Convert a financial burden into one normal device placed in one physically legible relationship with a tote or wallet.
0042  - Convert category rotation into one ordinary object being physically rotated while remaining visibly the same object.
0043  - Convert household spillover into one wrong recipient adapting with one simple household object.
0044  - Convert non-intervention into body orientation or furniture placement rather than cuffs, mitts, wrapped hands, or blocked doorknobs.
0045  - When a report requires relocation, choose a unique safe household setting and remove redundant protective gear.
0046  
0047  ## Deterministic validation
0048  
0049  Before delivery, parse the final file and verify:
0050  
0051  - exact top-level schema;
0052  - exact requested record count and IDs;
0053  - unique IDs and correct ad/candidate mapping;
0054  - required field names and field order when specified;
0055  - one person maximum per scene;
0056  - no prohibited motifs or banned visual dependencies in the actual concept text;
0057  - no files outside the requested deliverable were changed.
0058  
0059  Do not use a naive forbidden-term scan against QA explanations that merely name excluded motifs. Prefer positive generatability language in the final records so machine checks do not confuse exclusions with scene content.

Brand & doctrine / zakariah-copy-doctrine.md

/Users/ayden/.openclaw/workspace/knowledge/zakariah-copy-doctrine.md

SHA-256: e39567aa9123b190c2f33972f7c4761ab7307315d17173bf63e830895536235f

Full source · 221 lines
0001  ---
0002  domain: zakariah
0003  mode: copy-doctrine
0004  ttl: permanent
0005  load-when: every copy job, before writing anything
0006  ---
0007  # Zakariah — Copy Doctrine
0008  _The WHY and HOW before you write a single word. Read this first. Every time._
0009  
0010  ---
0011  
0012  ## THE FOUNDATION
0013  
0014  **Copy cannot create desire. It can only focus desire that already exists.** (Schwartz)
0015  Every word you write is aimed at desire that's already there. Your job is to focus it — not manufacture it.
0016  
0017  **Avatar research is the primary lever. Everything else is a multiplier on it.**
0018  Same product. Same price. Generic targeting → $68 CPA. Deep avatar → $16 CPA. The research IS the copy.
0019  
0020  **Format never wins without research. Research wins regardless of format.** (gulfam, $200K/day)
0021  
0022  ---
0023  
0024  ## THE 4-LAYER HIERARCHY (read before writing anything)
0025  
0026  Every piece of copy is built in this order. Never invert it.
0027  
0028  **Layer 1 — Doctrine (this file):** WHY you write what you write. Avatar, desire, awareness, mechanism, specificity. Foundation. Non-negotiable.
0029  
0030  **Layer 2 — Always-on style rules:** Every sentence earns the next. Short. Declarative. No run-ons. No padding. No filler words. No hedging. Applies to every format, every section, always.
0031  
0032  **Layer 3 — Format execution:** How doctrine is applied inside a specific container (native ad, advertorial, VSL, UGC). Read the relevant format file.
0033  
0034  **Layer 4 — Comp as quality ceiling, never template:** Extract principles from great copy, never phrasing. "The perimenopause series shows what's possible" — not what to copy.
0035  
0036  ---
0037  
0038  ## COPY EPISTEMOLOGY — THE META-RULES
0039  
0040  ### Specificity as Proof
0041  Specificity signals lived experience. Generality signals copy.
0042  
0043  **"The detail they wouldn't make up is the detail that converts."**
0044  
0045  The specific detail — the one that's almost embarrassingly precise — is what makes copy feel real. Anyone can write "she was upset." Nobody fakes "her daughter said, 'Mommy, are you okay?'"
0046  
0047  Rules:
0048  - Replace every vague emotional label with a specific physical scene
0049  - "She was struggling" → "She kept a towel on her pillow for eight months"
0050  - "He noticed" → "He looked down at their hands. Looked at her. Didn't say a word."
0051  - Specificity ratio: 3:1 specific vs. generic throughout the copy
0052  - Numbers over approximations: "47,342 customers as of last Tuesday" not "thousands"
0053  - Named details over categories: "148/92" not "high blood pressure"
0054  
0055  ### Gap vs. Trap (Mechanism Framing)
0056  The type of mechanism you choose determines the emotional activation ceiling.
0057  
0058  **Gap = Negligence:** The treatment failed to do what it never fully promised. Avatar can rationalize it away. "Maybe I just wasn't consistent enough." Emotional ceiling: frustration.
0059  
0060  **Trap = Rationalization-immune closed loop:** The treatment actively caused or worsened the condition it was supposed to fix. Avatar cannot escape this with logic — every attempt to fix it compounds the problem. Emotional ceiling: urgency, dread, trapped.
0061  
0062  Example: SSRI for perimenopause anxiety isn't just ineffective (Gap) — it suppresses dopamine, which is the one pathway she needs most (Trap). She can't rationalize taking it anymore.
0063  
0064  Use Trap framing whenever the mechanism allows it. It makes inaction feel physically dangerous, not just suboptimal.
0065  
0066  ### Comp Hierarchy as Discipline
0067  "Extract principles, not phrasing." This is a named rule, not just a vibe.
0068  
0069  When studying great copy:
0070  1. Identify the structural principle being used
0071  2. Identify the emotional mechanism being activated
0072  3. Write your own execution from scratch
0073  4. Never copy hooks, characters, or phrasing — always re-derive
0074  
0075  Drift happens when you copy structure without understanding doctrine. Doctrine is why it works. Structure is how it happened to work that time.
0076  
0077  ---
0078  
0079  ## THE 8 PSYCHOLOGICAL FRAMEWORKS
0080  
0081  **A. System 1 & System 2 (Kahneman)**
0082  - System 1 = fast, emotional, subconscious. This is who buys.
0083  - System 2 = slow, rational, analytical. This is who justifies.
0084  - Rule: Hook with emotion (System 1) → justify with logic (System 2)
0085  - Test: "Does this feel simple and emotional first... and smart after?"
0086  
0087  **B. Self-Determination Theory (SDT)**
0088  Three needs must be met before action:
0089  - Autonomy: "I'm choosing this myself." — no pressure, no hype
0090  - Competence: "I can do this." — make it feel simple and achievable
0091  - Connection: "This was made for me." — specificity, relatability
0092  
0093  **C. COM-B Framework**
0094  All three must be present for purchase:
0095  - Capability: Do they understand it and feel able to do it?
0096  - Opportunity: Is it available for them right now?
0097  - Motivation: Do they care enough to act? For low-salience products you must CREATE motivation through anticipated regret.
0098  
0099  **D. Narrative Identity & Emotional Transport**
0100  - Stories bypass resistance via mirror neurons
0101  - The reader must see themselves in the story
0102  - Generic emotional moments don't work. The story must feel stolen from the avatar's journal.
0103  
0104  **E. Processing Fluency**
0105  - If it takes effort to understand, they bounce. The brain reads "easy to process" as "must be true."
0106  - Short lines, short paragraphs, no jargon, easy flow, visual breaks.
0107  
0108  **F. Loss Aversion & Anticipated Regret**
0109  - Fear of loss > desire for gain (2:1 ratio). But raw fear feels manipulative.
0110  - Right way: story-embedded anticipated regret. "What if I wait too long?" inside the narrative, not as a direct command.
0111  - Rule: Make anticipated regret part of the story, not the pitch.
0112  
0113  ---
0114  
0115  ## 12 FOUNDATIONAL COPY RULES (Zakariah)
0116  
0117  1. **Customer First** — every sentence whispers "this person gets me"
0118  2. **Emotion first, logic second** — open emotional door, then facts walk them through
0119  3. **Specificity = believability** — "47,342 customers as of last Tuesday" not "thousands"
0120  4. **Tension creates movement** — reader feels uneasy until they resolve it by continuing
0121  5. **Sell the transformation, not the product** — product is just the vehicle
0122  6. **One reader, one message** — "mass copy = no copy"
0123  7. **Momentum is everything** — first line sells the second; every line earns its place
0124  8. **Readability > everything** — short sentences, visual breaks, nothing that makes you stumble
0125  9. **Story is the shortcut to belief** — one well-told moment beats 20 bullet points
0126  10. **Clarity is confidence** — vague = uncertain = kills conversions
0127  11. **Copy should sound like a conversation** — finish the sentence in their head
0128  12. **KISS (Granny-May Method)** — if tired/stressed/brain-dead scrolling could get it, you did it right
0129  
0130  ---
0131  
0132  ## COPY KILLERS — DELETE BEFORE FINALIZING
0133  
0134  **Words (delete immediately):**
0135  - "Game changer" / "life changing" / "revolutionary"
0136  - "Amazing" / "incredible" / "powerful"
0137  - Weak modifiers: "very," "really," "just," "truly"
0138  - Vague labels: "confident," "struggling" — replace with specific sensory moment
0139  
0140  **Structural killers:**
0141  - Writing at wrong awareness level (most common failure)
0142  - Wrong sophistication stage
0143  - Generic avatar — not specific enough to resonate with anyone
0144  - No villain — she blames herself → she freezes → no sale
0145  - CTA at end instead of at product reveal
0146  - LP doesn't match ad angle
0147  
0148  **Emotional killers:**
0149  - Trying to trigger all 6 emotions — pick ONE primary emotion, build to a peak
0150  - Blaming or shaming the avatar — always externalize the problem
0151  - Using fear as a lecture, not through story — story-embedded fear converts, lectures don't
0152  - Future pacing too vague — "you'll feel confident" vs. "you zip up the jeans in the back of your closet and your partner does a double take"
0153  - Generic testimonials — polish kills believability; Reddit/Amazon voice converts
0154  
0155  **Research killers:**
0156  - Generating or inferring quotes — only verbatim customer language
0157  - Isolated quotes without context
0158  - Broad avatars — "women who want to lose weight" is unsellable
0159  - Skipping avatar re-read before copy session
0160  
0161  ---
0162  
0163  ## THE VILLAIN REQUIREMENT
0164  
0165  **Never make them feel bad and it's their fault — always create a devil.**
0166  The villain is always external. Never the avatar.
0167  
0168  Without a villain, she blames herself → she freezes → no sale.
0169  With a villain, she has something to be angry at → anger creates motion → motion creates purchase.
0170  
0171  The villain can be: the system, the test, the wrong treatment, the pharmaceutical economics, the food industry, the doctor who missed it. Name it specifically. Let her fury land somewhere real.
0172  
0173  ---
0174  
0175  ## SCRIPT QA CHECKLIST (run before finalizing any copy)
0176  
0177  - [ ] Does the hook work at the right awareness level?
0178  - [ ] Does the avatar's exact language appear in the copy?
0179  - [ ] Is the mechanism specific, direct, and believable?
0180  - [ ] Is there ONE primary emotion being built to a peak?
0181  - [ ] Are there at least 3 open loops? Do they all close?
0182  - [ ] Is the specificity ratio 3:1 (specific vs. generic)?
0183  - [ ] Does the future pacing use time-stamped, sensory-specific moments?
0184  - [ ] Is there a CTA at product reveal (not just at the end)?
0185  - [ ] Does copy pass the "read out loud" test without stumbling?
0186  - [ ] Is the detail "the one they wouldn't make up"?
0187  - [ ] Is the villain external, never the avatar?
0188  - [ ] Does the copy satisfy COM-B? (Capability: understands mechanism. Opportunity: available now. Motivation: emotionally resonant.)
0189  - [ ] Does the copy satisfy SDT? (Autonomy: feels like her choice. Competence: simple enough. Connection: made for her.)
0190  
0191  ---
0192  
0193  ## 5 BEFORE/AFTER SENSORY MOMENTS
0194  
0195  **Before (the problem lived):**
0196  - Morning Moment — before she gets out of bed
0197  - Mirror Moment — what she sees and feels
0198  - Public Moment — when others might notice
0199  - Breaking Point — the moment she couldn't ignore it anymore
0200  - Alone Moment — the private thought she'd never say out loud
0201  
0202  **After (the transformation):**
0203  - Morning Moment (mirrored) — same scene, different feeling
0204  - Mirror Moment (mirrored) — same mirror, different person
0205  - Public Moment (mirrored) — the moment someone else noticed
0206  - Victory Moment — the specific win she wanted most
0207  - New Normal Moment — identity she got back
0208  
0209  ---
0210  
0211  ## KEY OPERATOR DOCTRINE (copy-specific)
0212  
0213  - *"Avatar research is the primary lever. Everything else is a multiplier on it."* (Zak)
0214  - *"The uglier the ad, the more native it looks. Ugly wins because it stands out."* (Zak)
0215  - *"Want purchase, not engagement."* Engagement metrics are noise. CPA is signal. (Zak)
0216  - *"Mechanism change is the biggest lever — bigger than creative, angle, or offer."* (Zak)
0217  - *"The moment you reveal the product — CTA goes there. Not at the end."* (Zak)
0218  - *"Super native, raw image, heavy copy — a super trojan horse ad."* (Suheyl)
0219  - *"Format never wins without research. Research wins regardless of format."* (gulfam)
0220  - *"I was breakeven for 3 months. Changed the mechanism. Hit $800K in the next 30 days."* (ceojachym)
0221  - Mechanism formula: new hope + easy to believe + backed by proof elements (ceojachym)

Brand & doctrine / brands/dawnbands/selling-root-cause-mechanism.md

/Users/ayden/.openclaw/workspace/knowledge/brands/dawnbands/selling-root-cause-mechanism.md

SHA-256: d4f3d47db06c7f53a54f18b79b181c09971609b8765bd1286bdad7c20a24fd53

Full source · 458 lines
0001  ---
0002  last_updated: 2026-08-10
0003  brand: dawnbands
0004  owner: fish
0005  status: canonical-selling-doctrine
0006  precedence: overrides older claim-safe-only mechanism framing for ad persuasion strategy; evidence/provenance gates still govern source attribution and fabricated testimonials
0007  ---
0008  
0009  # DawnBands — Canonical Selling Root Cause + Mechanism
0010  
0011  ## Mandatory Use
0012  
0013  Load this file before writing, briefing, editing, or auditing **any DawnBands ad copy, native, advertorial, UGC, VSL, listicle, lander, or mechanism section**.
0014  
0015  This is the selling doctrine.
0016  
0017  Older `root-cause.md`, `mechanism.md`, and claim-safe references remain research/substantiation inputs, but they do not flatten the consumer-facing mechanism into “different cue placement” alone.
0018  
0019  ## Operator Decision
0020  
0021  DawnBands copy may aggressively dramatize and simplify:
0022  
0023  - adolescent circadian delay
0024  - accumulated sleep debt
0025  - auditory gating / suppression of repeated room sound
0026  - sleep inertia and half-awake automatic behavior
0027  
0028  Use these as the biological selling layer because they make the failure of louder alarms believable and externalize blame from both Mom and the teen.
0029  
0030  The product connection must remain coherent:
0031  
0032  - biology explains why mornings are difficult
0033  - auditory gating + sleep inertia explain why sound alarms and task alarms fail
0034  - tactile wrist stimulation is the category-changing bypass
0035  - the Wrist-First Handoff explains the household and emotional payoff
0036  
0037  Do not weaken future copy back to feature-only language such as “sound and vibration are different.” Build the full causal chain.
0038  
0039  ---
0040  
0041  # The Root-Cause Stack
0042  
0043  ## Canonical name
0044  
0045  # The Deep-Sleep Signal Lock
0046  
0047  His morning is trapped behind four interacting biological layers.
0048  
0049  ### Layer 1 — Circadian Delay
0050  
0051  During adolescence, the internal sleep clock shifts later while school, work, sports, and family schedules continue demanding early waking.
0052  
0053  Selling frame:
0054  
0055  > His body is still protecting biological night at the exact hour the world demands consciousness.
0056  
0057  ### Layer 2 — Sleep Debt
0058  
0059  Short nights increase the pressure holding him asleep the next morning.
0060  
0061  Selling frame:
0062  
0063  > By morning, accumulated sleep pressure is pulling harder than the alarm is pushing.
0064  
0065  Aggressive frame:
0066  
0067  > By Friday morning, his brain is protecting sleep more aggressively than any ordinary room alarm can interrupt it.
0068  
0069  ### Layer 3 — Auditory Gating
0070  
0071  During deep sleep, repeated room sound can be pattern-matched, suppressed, or treated as background while the rest of the household hears it clearly.
0072  
0073  Selling frame:
0074  
0075  > His sleeping brain turns down repeated sound before it turns down touch.
0076  
0077  Aggressive frame:
0078  
0079  > The louder the room gets, the better his sleeping brain becomes at deleting it.
0080  
0081  Failed-solution trap:
0082  
0083  > Seven alarms do not create seven new chances to wake up. They create seven repetitions of the same signal his brain already filed under background noise.
0084  
0085  ### Layer 4 — Sleep Inertia
0086  
0087  The body can perform simple actions before full alertness, memory, and initiation return.
0088  
0089  This explains why he can:
0090  
0091  - silence the phone
0092  - answer Mom
0093  - sit up
0094  - walk across the room
0095  - solve a puzzle
0096  - return to bed
0097  - remember none of it later
0098  
0099  Selling frame:
0100  
0101  > The alarm wakes his hands without waking the part of him that can start the morning.
0102  
0103  ---
0104  
0105  # The ADHD Intensifier Layer (added 2026-08-06 — Fish directive; ADHD moms = current customer majority)
0106  
0107  ADHD adds no new problem. It turns every layer up, then adds one lock the general lane never named.
0108  
0109  - **Layer 1 hotter:** ADHD is strongly associated with delayed sleep timing (advertorial tier with citation: Van Veen et al. 2010; Kooij & Bijlenga 2013). Selling frame: *His clock isn't just teenage-late. It's ADHD-late on top of teenage-late.*
0110  - **Layer 2 hotter:** the racing mind that won't shut off at 11pm is the same brain that can't boot at 6:30am.
0111  - **Layer 4 hotter:** no-memory dismissals are a universal ADHD-community experience ("I set 47 alarms," "my body gets up and turns it off without me"). Frame as what ADHD families describe, never as clinical fact. **⚠️ 2026-08-10:** this covers the half-awake AUTOPILOT only. Do not extend it into an arousal-threshold claim — "ADHD teens are harder to wake by sound" is BANNED (unsupported; the one study on point trends the opposite way). The ADHD-specific layer is the Initiation Wall. See `dawnbands-avatar-adhd-teen-mom.md` Part 10.
0112  - **The Initiation Wall (the new lock):** even after consciousness, ADHD adults describe a wall between hearing the morning and starting it ("I physically cannot make my body move"). Waking up and getting up are two different problems; sound alarms address neither. Mechanism answer: the vibration builds and persists until he sits up and dismisses it — the dismissal is the first completed action of the day, within reach of a hand, not across a room. Identification framing only; never treatment.
0113  - **Household multiplier:** ADHD runs in the house. Dual-ADHD households (dad can't model mornings, mom runs wake service for two) are a live angle and the adult-version demand signal.
0114  
0115  Full doctrine, claim tiers, and intensification bank: `mechanism-fleshed-2026-08-06.md` §1a.
0116  
0117  ---
0118  
0119  # The Fifth Layer — Mom Becomes the Emergency Pathway
0120  
0121  When artificial signals are gated out or produce only half-awake behavior, Mom becomes the signal that finally breaks through.
0122  
0123  She knocks.
0124  
0125  She repeats herself.
0126  
0127  She opens the door.
0128  
0129  She escalates until emotion becomes novel enough to cut through.
0130  
0131  Core emotional line:
0132  
0133  > His brain ignores the alarms, but it notices the anger.
0134  
0135  This creates the memory asymmetry:
0136  
0137  > She remembers nine patient attempts. He remembers waking up to one angry mother.
0138  
0139  The biological problem becomes a relationship wound.
0140  
0141  ---
0142  
0143  # Complete Root-Cause Chain
0144  
0145  > delayed teenage clock → accumulated sleep pressure → repeated sound gets gated out → half-awake actions create false waking → Mom becomes the only signal that finishes the job
0146  
0147  ## One-paragraph reveal
0148  
0149  > His teenage body clock is still pulling him toward sleep when school demands he wake up. Sleep debt makes that pull even stronger. In deep sleep, his brain can suppress the same repeated room sounds while still letting his body perform automatic half-awake actions, which is why he can turn off five alarms, answer you, and remember none of it. When every artificial signal gets deleted, your escalating voice becomes the only novel signal left. His brain ignores the alarms, but it notices the anger.
0150  
0151  ---
0152  
0153  # The Failed-Solution Trap
0154  
0155  ## Canonical trap
0156  
0157  > Every louder alarm trains the household to depend more heavily on Mom while repeating the same signal pathway that already failed.
0158  
0159  The trap operates on two levels.
0160  
0161  ### Biological trap
0162  
0163  More alarms create more repetition inside the auditory pathway already being suppressed or pattern-matched as background.
0164  
0165  ### Household trap
0166  
0167  Every failed alarm recruits Mom as the rescue mechanism.
0168  
0169  The more reliably she rescues the morning, the more reliably the morning continues depending on her.
0170  
0171  Strong lines:
0172  
0173  > Every alarm in his room came with the same hidden backup feature: Mom.
0174  
0175  > You did not fail the alarm. The alarm kept failing back onto you.
0176  
0177  > The room got louder. The cue never became his.
0178  
0179  > Every morning you save may be teaching the system that it never has to work without you.
0180  
0181  Never blame Mom. She rescues the morning because the consequences are real. The villain is the system that leaves her no safe way to step out.
0182  
0183  ---
0184  
0185  # Unique Mechanism Stack
0186  
0187  ## Biological mechanism name
0188  
0189  # The Somatic Wake Bypass™
0190  
0191  ## Household mechanism name
0192  
0193  # The Wrist-First Handoff™
0194  
0195  Use both. They answer different resistance.
0196  
0197  ### Somatic Wake Bypass™
0198  
0199  DawnBands places a tactile signal directly against the wrist, using a sensory pathway outside the repeated room-sound pattern.
0200  
0201  Core explanation:
0202  
0203  > Sound has to cross the room and survive his brain’s filter. DawnBands starts past the filter.
0204  
0205  Supporting lines:
0206  
0207  > Stop asking the muted pathway to work harder.
0208  
0209  > Instead of making the room louder, put the wake signal on his body.
0210  
0211  > A loudspeaker shouts to everyone and hopes the right person responds. A wrist cue starts against the person who needs it.
0212  
0213  > The answer is not another sound for his sleeping brain to delete. It is a signal delivered through touch.
0214  
0215  ### Wrist-First Handoff™
0216  
0217  The tactile cue begins on the teen before Mom’s voice enters the morning.
0218  
0219  Core explanation:
0220  
0221  > The first cue starts on his wrist before it ends in your voice.
0222  
0223  Supporting lines:
0224  
0225  > Not louder. On him.
0226  
0227  > Move the first signal from your throat to his wrist.
0228  
0229  > The goal is not a better tenth attempt. It is a first cue that does not require Mom.
0230  
0231  > DawnBands does not make Mom better at waking him. It removes her from the first step.
0232  
0233  ### The Ownership Trade™ (added 2026-08-06)
0234  
0235  Third named mechanism. Answers wear-doubt (15% of surveyed buyers: "he'll take it off").
0236  
0237  The teen hates the current system more than she does: woken by the angriest version of his mother, dragged out of bed like a child in the house where he's fighting to be treated like an adult. His desire and hers are the same desire — get mom out of the morning.
0238  
0239  Core explanation:
0240  
0241  > He doesn't wear it for the alarm. He wears it to get you out of his morning.
0242  
0243  Why the trade holds: it's his, and it reports nothing to anyone (structurally incapable of surveillance); no screen to out him at school or sleepovers; silent, so waking stops being a household event; and the alternative is mom.
0244  
0245  Copy rule: wear-objection handling is TEEN-motive framing, never compliance framing. "Make him wear it" resurrects surveillance and kills the trade.
0246  
0247  Full logic: `mechanism-fleshed-2026-08-06.md` §1b.
0248  
0249  ---
0250  
0251  # Full Selling Logic Chain
0252  
0253  ## Root cause
0254  
0255  > His delayed teenage clock and accumulated sleep debt pull him into deeper morning sleep.
0256  
0257  ## Hidden mechanism
0258  
0259  > In that state, his brain gates out repeated room sound while leaving him capable of automatic half-awake actions.
0260  
0261  ## Failed-solution mechanism
0262  
0263  > More alarms repeat and reinforce the same sound pattern his sleeping brain already suppresses.
0264  
0265  ## Emotional consequence
0266  
0267  > When every artificial signal gets deleted, Mom’s escalating voice becomes the only novel signal left.
0268  
0269  ## Product mechanism
0270  
0271  > DawnBands bypasses the muted auditory pathway with a direct tactile signal on the wrist.
0272  
0273  ## Household mechanism
0274  
0275  > The first wake cue belongs to the teen before Mom has to enter the room.
0276  
0277  ## Functional payoff
0278  
0279  > The morning gets a chance to begin on his body instead of ending in her throat.
0280  
0281  ## Emotional payoff
0282  
0283  > He stops waking up to the angriest version of her.
0284  
0285  ## Future payoff
0286  
0287  > The first evidence that his morning can belong to him before college, work, or a roommate inherits the job.
0288  
0289  ---
0290  
0291  # Analogy System
0292  
0293  ## Primary biological analogy — Muted Channel
0294  
0295  > Imagine turning up a television after someone muted the input. More volume does not matter if the pathway carrying the signal has already been shut down. Sound alarms keep turning up the room. DawnBands changes the pathway.
0296  
0297  ## Primary household analogy — Relay Handoff
0298  
0299  > Waking is a relay, not a light switch. The alarm runs the first leg. When it drops the baton, Mom picks it up and finishes the handoff. DawnBands moves that first pass onto his wrist before she has to enter the race.
0300  
0301  ## Short analogy — Loudspeaker Versus Tap
0302  
0303  > A room alarm is a loudspeaker. DawnBands is a tap delivered directly to the person who needs the signal.
0304  
0305  Use one analogy per execution. Do not stack all three in one ad.
0306  
0307  ---
0308  
0309  # Mechanism Delivery by Format
0310  
0311  ## Long problem-aware native
0312  
0313  Use the full stack:
0314  
0315  1. lived morning wound
0316  2. delayed clock + sleep pressure
0317  3. auditory gating
0318  4. half-awake sleep-inertia behavior
0319  5. Mom as emergency pathway
0320  6. tactile bypass
0321  7. Wrist-First Handoff
0322  8. emotional/future payoff
0323  
0324  ## Short native
0325  
0326  Compress to:
0327  
0328  > His brain was not ignoring every alarm because he did not care. In deep sleep, repeated room sound was the pathway already failing. DawnBands puts the signal on his wrist before my voice has to become the alarm again.
0329  
0330  ## Objection killer
0331  
0332  > “He sleeps through everything. Alarms right next to his head. Nothing gets through.” Fair. The mechanism is not “louder finally works.” Every alarm she has tried travels the same gated sound channel. The test is whether a direct cue worn on his wrist arrives on a channel his sleeping brain does not gate at all.
0333  
0334  Then use exact-SKU proof when available.
0335  
0336  ⚠️ Objection-killer copy rule (Fish 2026-05-14 + 2026-07-19, 3rd doc hit fixed 2026-07-22): NEVER name bed shakers, vibrating pillows, or any tactile failed solution in copy — same-channel confound with our mechanism. Failed solutions in copy are SOUND, LIGHT, ROUTINE, or APP based, genericized (no competitor brand names), with prices kept. Full rule: `failed-solutions-graveyard.md` line 11.
0337  
0338  ## UGC
0339  
0340  Use casual language:
0341  
0342  > Apparently his body can turn an alarm off before the part of his brain that forms a memory is actually online. That explained everything. The band does not make the room louder. It puts a different signal right on his wrist.
0343  
0344  ## Advertorial/VSL
0345  
0346  Reveal each biological layer gradually, then make the wrist cue feel inevitable. (2026-08-06: do NOT print the internal ™ names in the ad — deliver the three-part mechanism in plain language per Rule 10.)
0347  
0348  ## Listicle/product-aware
0349  
0350  Lead with the distinction:
0351  
0352  > Sound has to cross the room and survive his brain’s filter. DawnBands starts past the filter.
0353  
0354  ---
0355  
0356  # Required Future-Copy Rules
0357  
0358  1. Use the biological root-cause stack as the default selling doctrine for DawnBands.
0359  2. Do not flatten the mechanism to “vibration is different from sound.” Explain why sound failed.
0360  3. Keep circadian delay and sleep debt as severity amplifiers that deepen the Signal Lock.
0361  4. Make auditory gating the primary product-congruent biological root cause.
0362  5. Use sleep inertia to explain automatic alarm dismissal, false waking, and missing memory.
0363  6. Make Mom becoming the emergency pathway the emotional and household consequence.
0364  7. Use Somatic Wake Bypass™ for the biological solution.
0365  8. Use Wrist-First Handoff™ for ownership, independence, and relationship payoff.
0366  9. Always externalize blame from both Mom and the teen.
0367  10. Product reveal should feel like the inevitable answer to the pathway failure, not a random wristband.
0368  11. Exact-SKU superiority, buyer timelines, testimonials, and product outcomes still require receipts. Aggressive mechanism framing does not authorize invented customer histories.
0369  12. Preserve native voice. The narrator should not suddenly sound like a neuroscientist unless the format provides a believable authority/discovery bridge.
0370  
0371  ---
0372  
0373  # Claims and Provenance Separation
0374  
0375  This doctrine intentionally distinguishes two jobs.
0376  
0377  ## Selling layer
0378  
0379  Aggressive, simplified biological explanation designed to create belief and new hope:
0380  
0381  - delayed clock
0382  - sleep pressure
0383  - auditory gating
0384  - sleep inertia
0385  - tactile bypass
0386  
0387  ## Substantiation layer
0388  
0389  Internal evidence and review used to prevent fabricated proof:
0390  
0391  - current product specifications
0392  - exact-SKU behavior
0393  - attributed research
0394  - same-speaker testimonials
0395  - current price/refund terms
0396  
0397  The substantiation layer must not neuter the selling mechanism into generic feature language.
0398  
0399  The selling layer must not invent customer outcomes, exact product performance, medical treatment claims, or fake receipts.
0400  
0401  ---
0402  
0403  # Canonical Copy Spine
0404  
0405  > private morning rupture → biological innocence → Deep-Sleep Signal Lock → louder-alarm trap → half-awake behavior → Mom becomes emergency pathway → Somatic Wake Bypass™ → Wrist-First Handoff™ → desired absence → independence/future stakes → proof/offer
0406  
0407  # Final Standard
0408  
0409  Every future DawnBands copy should make the reader understand:
0410  
0411  1. why his biology is pulling him toward sleep
0412  2. why sound alarms keep failing
0413  3. why his half-awake actions are not proof he was consciously awake
0414  4. why Mom keeps becoming the final signal
0415  5. why a tactile cue on the wrist is a new pathway, not another louder alarm
0416  6. why moving the first cue onto him changes more than noise
0417  7. why this matters for their relationship and his future independence
0418  
0419  If the copy cannot answer all seven where format length permits, the mechanism is underbuilt.
0420  ---
0421  
0422  # Required Future-Copy Rules (added 2026-07-22 — binding guardrails, previously scattered in sibling docs)
0423  
0424  1. ~~ADHD identification only — ONE identification line cap~~ **STRUCK 2026-09-02: the one-line cap governed the generic lane, which is retired. Rule 8 governs: ADHD is named in the hook.** What survives of rule 1: never efficacy/treatment claims tied to ADHD. (Fish 2026-07-19; full must-reject list in `mechanism.md` AVOID section — that list remains binding even though mechanism.md is research-subordinate.)
0425  2. **Failed solutions genericized.** No competitor brand names in copy ("the famous extra-loud alarm clock," "smart speakers," "the alarm app," "a smartwatch"). NEVER bed shakers / vibrating pillows / any tactile failed solution — same-channel confound. Failed stacks are SOUND, LIGHT, ROUTINE, APP based. Prices stay and should total more than product price. (Fish 2026-05-14 + 2026-07-19; `failed-solutions-graveyard.md` L11.)
0426  3. **Lane scoping (amended 2026-09-02).** ONE lane: mother of a 13–18 ADHD teen. Parents-of-8-12, generic deep-sleeper, suspecting-undiagnosed, ADHD adult, partner, deaf/HoH are all out of brief scope until Fish reopens them. Vary teen age 13–18 and gender across every batch.
0427  4. **Format rulings 2026-07-22:** ALL-CAPS hooks allowed. Rich/emoji trust footers allowed. No word-count caps in any mode — vary lengths deliberately. Em-dash ban unchanged.
0428  5. **Canonical failed-solution stack (survey-locked, refreshed to n=138 on 2026-09-02):** waking him myself (89%) → phone alarms full volume (80%) → loud alarm clock (55%) → routines/bedtimes/sleep apps (50%) → a smartwatch (16%, downgraded from mandatory kill to optional beat). Bed shaker 13% — research only, banned in copy. Items 1-2 are the anchor tier. The routines/apps beat is MANDATORY in ADHD-lane copy (blame removal: the effort was real, the target was wrong). Sunrise clock optional sixth. Bed-shaker ban unchanged. Per-item mechanism kills: `mechanism-fleshed-2026-08-06.md` §1c.
0429  6. **Smartwatch kill is non-channel only:** calibration (a notification is engineered to be ignorable), saturation (habituation scoped to meaningless all-day repetition INSIDE a channel, never to "tactile signals" as a category), surrender (fires briefly, sleep modes mute it, wrist-raise hands it to the autopilot, battery dies), screen-in-bed. Any line implying touch itself habituates is banned — it hands the reader "so he'll tune this out too."
0430  7. **Guarantee adjacency (Fish-approved direction 2026-08-06):** every mechanism section carries "100% wake-up guarantee. 60 nights." in the same breath as the wake claim — it answers the #1 buyer hesitation (50%). The full trio with $49 is reserved for the close/CTA, after value is built.
0431  8. **ADHD lane directness (Fish 2026-08-06 — amends rule 1):** the dedicated ADHD lane may name ADHD in hooks and headlines and explain ADHD-morning biology at the tiered claim levels in `mechanism-fleshed-2026-08-06.md` §1a. The general lane keeps the one-identification-line cap. Efficacy and treatment claims remain banned in ALL lanes: no "treats/cures ADHD," no "helps ADHD symptoms" as a category, no medication-replacement or reduce/stop-medicating claims, no medical/FDA language. ALLOWED and encouraged — villainize the treatment GAP: no pill can work on a kid who is still asleep; the prescription never covered 6:30am; she has been the unprescribed step between the bottle and his mouth (Invisible Bridge). The product fixes the hour treatment never touched. It never touches the treatment itself.
0432  9. **Humanizer pass is mandatory for ALL consumer-facing copy, template or batch (Fish 2026-08-06):** draft → humanizer (golden TrendTrack native/static copies from proven brands + voc-syntax-bank + phrasing-bank as reference corpus) → qa-gate. Analogies come from the avatar's daily world (fridge hum, background noise, grabbing his shoulder); if a metaphor needs a beat of decoding, use the plain version. qa-gate flags a missing humanizer changelog as BLOCKING.
0433  10. **Mechanism names are INTERNAL ONLY (Fish 2026-08-06):** the ™ names (Deep-Sleep Signal Lock, Somatic Wake Bypass™, Wrist-First Handoff™, Ownership Trade™) never appear in consumer copy. Copy delivers the three-part mechanism in plain language: (1) touch starts past the sound filter, (2) it builds and doesn't stop until he turns it off awake, (3) the first signal starts on him, not mom. This supersedes the earlier Advertorial/VSL instruction to "name the Deep-Sleep Signal Lock and Somatic Wake Bypass™" — internal docs keep the names for routing; ads never print them.
0434  11. **Wear beat is mandatory in every piece (Fish 2026-08-06):** every ad/lander/mechanism section carries at least a brief ownership line answering "would he actually wear it" — teen-motive framing (he keeps it on because it gets mom out of his morning; silent, not a phone or smartwatch, reports to nobody, nothing to out him). **"No screen" is banned as of 2026-08-27 — the band has an LED time readout (`product-truth.md`).** One or two lines minimum; full trade logic on landers.
0435  12. **Brand name in copy is "Dawn Bands" — two words, always (Fish 2026-08-06).** "DawnBands" one-word stays only in internal docs, file names, and code.
0436  13. **Substantiation is governed by `substantiation-2026-08-06.md` (verified 2026-08-06).** Two changes bind immediately:
0437      - **Layer 3 re-anchor.** Stop leading auditory gating with "his brain got used to HIS alarm over months" — that specific claim is unverified extrapolation, and the nearest evidence (Perrin et al. 1999) shows personally-meaningful sound wakes people FASTER. Lead instead with arousal threshold, which is well replicated and more dramatic: over half of children tested do not wake to a smoke alarm engineered to save their life (Pediatrics 2022, N=540; Smith 2006, 58% woke to a tone alarm). Keep "the brain tunes out repeated background sound" as a general, true statement. Never "your alarm wore out."
0438      - **New Layer 5 proof.** Smith et al., Pediatrics 2006: a recorded mother's voice woke 96% of children vs 58% for a standard tone alarm, and roughly ten times faster. This is scientific proof of her lived reality — the generic alarm failed and she worked. Land it on the household consequence (she was never the backup, she was the only thing that worked), never on acoustics, or it argues for a voice alarm instead of the band.
0439  14. **Gentle-wake angle (Fish 2026-08-06) — narrow version only, secondary beat, 1-2 lines max, never the focal point.** ALLOWED, hedged verbatim: "some research suggests being startled awake produces a bigger blood-pressure and heart-rate jump than waking up gradually" (Kim et al. 2024; Kaida 2005; Hall 2016 — all small). ALLOWED, brief cortisol clause, one line, no numbers, placed after the wake-doubt work is done: "researchers who measured it found a nighttime alarm pushed up heart rate and stress hormones" (Hall et al., Noise & Health 2016, N=16). ALLOWED unhedged as product fact and the preferred default: it wakes him without a sound going off in the room, no siren, no jolt, nothing waking the house. BANNED: CAR framing ("waking spikes your cortisol and that's bad for you" — contradicted by Klaas 2025), any cortisol number or percentage, WHO noise guidelines, "bad for your heart," and any claim that vibration is proven calmer or lower-stress than sound (no such study exists; a competitor publishes that claim uncited — do not echo it). Full kill list and the CAR-vs-acute distinction: `substantiation-2026-08-06.md`.
0440  
0441  # "That's Why" Map (symptom → chain link → causal logic)
0442  
0443  The conversion engine: after the mechanism reveal, re-explain HER OWN symptoms through the chain. Voice each line fresh per ad in the narrator's register (never copy these — causal logic only, per Fish's function-not-lines rule):
0444  
0445  | Her observed symptom (from VOC) | Chain link | Causal logic to voice fresh |
0446  |---|---|---|
0447  | Sleeps through 3+ alarms, fire alarms, smoke detectors | Auditory gating / Deep-Sleep Signal Lock | Volume never mattered — every alarm rides the same gated sound channel; his brain closes that door in deep sleep |
0448  | "He answers me, then swears I never woke him" / turns alarms off with no memory | Sleep inertia / half-awake automatic behavior | He wasn't lying and wasn't awake — his body acts before the memory-forming brain is online |
0449  | Walks to the phone across the room, returns to bed, eyes open | Same | Automatic behavior defeats distance tricks; relocation is still the same channel + an unconscious body |
0450  | Louder alarms work for a week, then die | Habituation on the gated channel | Escalating volume trains the gate to work harder — the loudest alarm fails fastest |
0451  | Only Mom's shake works | Mom = emergency pathway | Touch is the one channel his brain never gates — she has been the tactile alarm all along |
0452  | Weekends fine, school days impossible | Circadian delay + sleep debt | Saturday isn't discipline — school time sits hours before his shifted biological morning |
0453  | Wakes exhausted even after "enough" hours | Sleep debt + inertia | The debt collects interest; the wake-up fight spends the first hour of his day |
0454  | She wakes before her own alarm, dreading his | Mom-as-pathway cost | Her body learned his wake-up is her job — the anticipation is the pathway running in reverse |
0455  | "He has a smartwatch, it taps his wrist" | Smartwatch saturation/calibration kill | A notification device delivers notifications: engineered to be dismissible, buried under forty identical buzzes a day, gives up before deep sleep does |
0456  | "Even when he's awake he won't get out of bed" | Initiation Wall (ADHD lane) | Waking and starting are two different problems — the band's build-and-persist makes ending it the first completed action of the day (identification framing only) |
0457  
0458  Copy shape: 2-4 "that's why" beats per ad maximum, chosen to match the symptoms the ad's opening actually planted. Every beat must close a loop the reader recognizes from her own mornings.