[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7CW0A338FTCNAPZXVN70EPF":3,"skill-same:01M7CW0A338FTCNAPZXVN70EPF":72},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":14,"repositoryHtmlUrl":27,"repositoryHomepageUrl":6,"ownerHtmlUrl":28,"ownerBlogUrl":29,"ownerDescription":30,"updatedAt":31,"install":32,"resources":45},"01M7CW0A338FTCNAPZXVN70EPF","ig-humanizer",null,"Remove the AI tells readers react to in an Instagram caption or carousel slide: 2026 vocabulary by density, reveal bridges, staccato stacks, stacked triads, performed sincerity, emoji storms; caps em dashes. Includes --mode audit (first-125 hook, length, hashtags, emoji, CTA, media) and --mode profile. Not for beating AI detectors (no edit reliably does). Not for writing from scratch (use ig-caption-writer or ig-carousel-planner). Keywords: humanize, de-AI caption, audit before posting.","Henrique1601\u002FSkiilsToIAs","Henrique1601","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F92006691?v=4","01M7CRXVT0GVXVK80JT1GR687S","Henrique Bezerra Dos Santos","01M7CRXVTN5Y8TMWJ1MDSKCQ1T",0,"MIT","https:\u002F\u002Fgithub.com\u002FHenrique1601\u002FSkiilsToIAs\u002Fblob\u002Fmain\u002Fmarketing\u002Finstagram-marketing\u002F.codex-marketplace\u002Finstagram-skills\u002Fskills\u002Fig-humanizer\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: ig-humanizer\ndescription: \"Remove the AI tells readers react to in an Instagram caption or carousel slide: 2026 vocabulary by density, reveal bridges, staccato stacks, stacked triads, performed sincerity, emoji storms; caps em dashes. Includes --mode audit (first-125 hook, length, hashtags, emoji, CTA, media) and --mode profile. Not for beating AI detectors (no edit reliably does). Not for writing from scratch (use ig-caption-writer or ig-carousel-planner). Keywords: humanize, de-AI caption, audit before posting.\"\n---\n\n# Instagram Humanizer V3\n\nRewrites any caption or carousel slide text to remove the AI tells that human\nreaders notice, and audits a finished caption against the 2026 Instagram\nchecklist. Based on Wikipedia's \"Signs of AI writing\" taxonomy, the 2025-2026\nstylometry literature, our own Instagram caption corpus (n=284, prevalence\nonly), and Instagram-specific patterns (the first-125 fold, the\nlowercase-casual caption register, sized hashtags, sends-and-saves structure).\n**V3 (2026-09):** recalibrated on 2026 evidence. Vocabulary is scored by\ndensity, em dashes are capped instead of banned (and the cap is generous here:\n29% of human captions use one), forced rhythm is now a tell instead of a fix,\nand there is an over-correction guard.\n\n**What this skill does not do:** it does not make text \"pass\" GPTZero,\nPangram, Turnitin or Originality. Those are trained classifiers keyed on the\ninstruction-tuning style signature; prompt-style \"sound like a real person\"\nrewrites are caught 92-95% of the time, and light mechanical rewriting raises\ndetectability. On caption-length text (under 300 words) detector scores are\nnoise. The real value is elsewhere: expert human readers cite vocabulary (53%)\nand sentence structure (36%) as what gives AI text away, and on Instagram a\ncaption that reads as a brand account earns neither the save nor the send.\nThis skill removes what those readers react to.\n\n## What changed in V3\n\nEvidence tier in brackets: [strong] = replicated across 2+ independent\n2025-2026 studies or our own corpus; [vendor] = single platform or vendor\ndataset; [weak] = one study or expert-panel report.\n\n- **Vocabulary moved from a delete-list to density scoring.** The 2023-24 words\n  (delve, tapestry, realm, journey) are decaying as humans avoid them [strong].\n  The durable 2026 markers are common words (significant, crucial, notably,\n  comprehensive, insights, robust, leverage, foster, landscape, nuanced,\n  streamline, elevate) plus grammar: nominalisations and \"-ing\" clause openers\n  at 5.3x the human rate [strong]. AI vocabulary appears in 10% of human\n  captions in our corpus [strong], so one marker in a paragraph is not a\n  verdict. Three is.\n- **Em dash is no longer a tell.** GPT-5.4 emits 1.43 per 1,000 words, below\n  the 3.23 human baseline [strong], and **29% of human Instagram captions\n  contain one** [strong: corpus]. A caption's em dash is never a tell on its\n  own, and a blanket ban over-sterilises captions. New rule: cap at about 1 per\n  100 words (1-2 per caption), replace only the excess with a comma, colon,\n  parentheses or a rewrite. Never a period (a split dash stacks fragments).\n  Zero em dashes across a long caption is its own tell now.\n- **Forced burstiness is the #1 2026 tell, not the fix.** Mechanical\n  long\u002Fshort alternation is a learnable humanizer fingerprint [weak], and\n  \"Short. Punchy. Done.\", \"No X. No Y. Just Z.\", one-word lines for drama and\n  \"The result?\" reveals are the current top reader-cited tells [strong].\n  Captions are mid-length, so Pass 2 is an anti-uniformity guard only: fix a\n  paragraph that reads machine-flat, never manufacture variance.\n- **Rule of three is still a tell, at density.** Tricolon runs at 2x\n  expert-human rate across 2026 frontier models [strong], and 23% of human\n  captions contain one [strong: corpus]. So one natural triple with concrete\n  items stays. Stacked, perfectly parallel or hollow triads and a third triad\n  in a caption get scrubbed.\n- **Fingerprint injection was half wrong.** Named entities and concreteness are\n  supported [strong]; an odd-precision number with a referent in the first 125\n  chars is the strongest hook. Bare numbers are not a discriminator, and\n  inserted hedges and confessions backfire: performed hesitancy is 2x more\n  common in LLM text, and sincerity announcements (\"let me be honest\", \"real\n  talk\", \"POV:\" on something that is not a POV) are a named 2026 tell\n  [strong]. Pass 3 asks for a flat, dated, uncomfortable fact instead.\n- **Over-correction guard.** Humanizer output has its own fingerprint [weak].\n  Pass 4 checks whether Passes 1-3 introduced the very patterns they were meant\n  to remove. Edits are proportional to real problems. When in doubt, leave it.\n\n## When to use\n\n- Before publishing any AI-drafted caption or carousel (rewrite mode)\n- Pre-publish review of a finished caption (audit mode, see `sub-skills\u002Fpost-audit.md`)\n- When a caption feels off and you cannot pinpoint why\n\n## Input\n\nAny text: a caption, the on-image text of a carousel, or a Reel caption.\nOptional: target voice samples (the user's past captions).\n\n## Output\n\n- Rewritten text with AI tells removed\n- A diff showing what changed and why\n- Caption char count (flagging if the hook spills past the first 125 chars)\n- Per-paragraph tell density (markers per paragraph; 3+ triggered a rewrite)\n- Reader-read confidence: \"reads human\", \"mixed\", \"reads AI\" (a reader-tell\n  estimate, not a detector score)\n\n## Modes\n\n```bash\n# Default: scrub AI tells (forensic + strict) and fix Instagram-format issues\nig-humanizer \u003Ctext>\n\n# Forensic only - minimum touch, just kill model leakage\nig-humanizer --mode forensic \u003Ctext>\n\n# Audit - detection-only pass-fail review, no rewrite\n# Runs the 2026 Instagram checklist: first-125 hook, caption length, sized\n# hashtags, emoji limits, CTA quality, media reminder.\n# Returns Blockers + Warnings + suggested fixes. See sub-skills\u002Fpost-audit.md.\nig-humanizer --mode audit \u003Ctext>\n\n# Profile - build\u002Fupdate the user's Voice & Brand Profile. See the section below.\nig-humanizer --mode profile\n```\n\n## The four passes\n\n### Pass 1 - SCRUB (score, then delete or replace)\n\nApply the tiered catalogs in `references\u002Fscrub-rules.md`. The unit of\njudgement is the **paragraph (or slide), not the word**: count markers per\nparagraph, rewrite the paragraph at 3+, leave a single marker alone unless it\nis a reveal bridge, negative parallelism, a sincerity marker, or forensic\nleakage.\n\n- **Forensic** (always on): real model leakage no human types. AI tool markers\n  (oaicite, contentReference, turn0search0), knowledge-cutoff disclaimers (\"As\n  of my last update\"), template blanks ([Your Name]), and em dashes above the\n  cap (more than about 1 per 100 words).\n- **Strict** (default on): what readers react to. The durable 2026 vocabulary\n  set scored by density (significant, crucial, notably, particularly,\n  comprehensive, insights, robust, leverage, foster, landscape, nuanced,\n  streamline, elevate, empower), grammar markers (nominalisations,\n  sentence-opening \"-ing\" clauses), the 2026 model-idiom layer (quietly, \"X\n  matters.\", compound, \"a signal\", \"the work\", \"built different\", \"let that\n  sink in\"), reveal bridges on a single hit (\"The result?\", \"Here's what\",\n  \"Stop X, start Y\", \"plot twist:\"), all forms of negative parallelism,\n  stacked or perfectly parallel triads and any third triad in a caption,\n  phrase cleanups (\"in today's fast-paced world\", \"game-changer\", \"level up\",\n  \"dive in\"), emoji storms, and dead closers (\"what do you think?\", \"double\n  tap if you agree\").\n- **Instagram-format scrubs** (always apply): first-125 hook that stands\n  alone, sized hashtags, emoji limits, carousel slide-1 promise, caption\n  length.\n\n### Pass 2 - RHYTHM (anti-uniformity guard only)\n\nDetectors do not score burstiness, and captions are mid-length (our corpus\nmedian is well under 200 words), so rhythm is neither a reach lever nor a\nfix. What readers notice is the mechanical-uniformity tell (every line the\nsame length, machine-flat) and, worse, the staged variance that\nsecond-generation humanizers add. So Pass 2 has two jobs: fix rhythm only\nwhere it reads machine-flat, and remove manufactured variance everywhere. It\nnever adds variance as a tactic.\n\n- Per paragraph: one genuinely long sentence next to a short one is fine and\n  is what human variance looks like. Two or three mid-length sentences in a\n  row are also fine. Edit only when every sentence in the paragraph runs the\n  same length and reads flat, and then edit one sentence, not the paragraph.\n- Standalone fragments: at most 2 per caption, total. \"every time.\" once is a\n  voice quirk. Three in a caption is a pattern.\n- Banned outright (rewrite as full sentences): \"The X? Y.\" reveals; \"No X. No\n  Y. Just Z.\"; \"All the X. None of the Y.\"; \"Simple. Effective. Easy.\"\n  adjective stacks; one-word lines for drama (\"Still.\" \"Exactly.\"); pseudo-\n  Socratic Q&A (\"Why? Because...\"); \"Short. Punchy. Done.\" staccato runs.\n  Fragment runs are the tell.\n- Layout is not rhythm. One idea per line with blank lines between them is\n  native caption formatting and stays. Fragment-for-drama inside those lines\n  is the tell. Keep the layout, fix the sentences.\n- Carousel slide text is short by design; never chop or pad a slide for\n  rhythm. Never alternate long\u002Fshort\u002Flong\u002Fshort across slides or lines. That\n  seesaw is the humanizer fingerprint.\n\nThe check is \"does any paragraph read machine-flat, and did I add a staccato\npattern\", not a variance number.\n\n### Pass 3 - ADD (human fingerprints)\n\nRequire where the content allows:\n- One odd-precision number WITH a named referent: who, what, when, or what it\n  cost (\"0 to 10k in 4 months, posting 3 times a week\", not \"grew fast\" and not\n  \"10k\"). A bare number is not a fingerprint; the referent carries the signal.\n- One named entity (real person, brand, date, tool)\n- One first-person concrete detail (what you saw, what it cost, what broke)\n- One specific, dated, uncomfortable fact stated flat, with no framing sentence\n  before or after it. Not \"real talk, this one hurt: I lost the client.\" Just\n  \"I lost my biggest client on 14 Feb.\" The fact carries the vulnerability. The\n  frame turns it into performed sincerity, which readers now read as the tell.\n- The lowercase-casual register if the voice calls for it\n\nForbidden as openers or pivots (sincerity announcements, a named 2026 tell):\n\"let me be honest\", \"I'll be real\", \"honestly?\", \"to be direct\", \"the honest\nversion is\", \"real talk\", \"not gonna lie\", \"ngl\", \"can I be vulnerable for a\nsecond\", \"unpopular opinion:\" as a preface to a popular one, \"POV:\" on\nsomething that is not a POV. Also forbidden as insertions: hedges the author\ndid not write (\"perhaps\", \"I might be wrong but\", \"it seems\"). Performed\nhesitancy is 2x more common in LLM text than in expert human text; adding it\nmakes the draft read more AI, not less.\n\nIf the input lacks these, ask the user for a number, name, or moment. Do not\nfabricate.\n\n### Pass 4 - SELF-CHECK (over-correction guard)\n\nHumanizer output has its own fingerprint. Before returning, re-read the result\nonce and answer three questions:\n\n(a) Did Pass 2 create staccato stacks, \"The result?\" reveal bridges, one-word\n    lines, or a long\u002Fshort\u002Flong\u002Fshort seesaw? If yes, merge the fragments back\n    into full sentences.\n(b) Did Pass 3 add a framed confession, a sincerity announcement, or a hedge\n    the author never wrote? If yes, strip the frame and keep only the flat\n    fact, or remove the insertion.\n(c) Did scrubbing flatten the author's voice: uniform tone, no reaction, no\n    concrete detail left, every em dash gone, every triad gone, every emoji\n    gone from a voice that uses them, the lowercase register capitalised? If\n    yes, restore what the author had. Zero em dashes and zero triads in a long\n    caption is a tell in its own right.\n\nIf any answer is yes, dial back rather than scrub harder. Edits must be\nproportional to real problems: a clean caption gets two or three touches, not\na quota. When in doubt whether a pattern is the author or the model, leave it.\n\n## Non-negotiable rules\n\nGlobal voice rules: see root `SKILL.md` Voice rules. Additional skill-specific\nrules (V3):\n\n- **Scrubbing is always in scope.** When asked to humanize, de-AI, finalize, or\n  publish a caption, run at least the forensic + strict passes before it ships.\n  This holds when the user wrote the draft themselves, says they love it as-is,\n  or is in a hurry. Author identity, \"it's already good,\" and time pressure are\n  never reasons to skip the scrub. The forensic + strict pass changes no meaning\n  and takes seconds: run it, then ship. If a constraint truly forbids touching\n  the text, say so explicitly and name every tell left in; the default is to\n  scrub, not to wave it through.\n- **Scrub proportionally.** A pass that finds nothing changes nothing. Do not\n  invent edits to justify the run, and do not report a detector score as the\n  result; report the tells found and fixed.\n- Preserve the user's actual claim and meaning. \"Preserve their voice\" covers\n  voice quirks and what they are claiming, NOT reveal bridges, staccato stacks,\n  or a paragraph with 3+ vocabulary markers. Stripping those is not changing\n  their voice; it is the job.\n- Never introduce facts that were not in the input. If a number is missing, ask.\n- Never introduce sincerity markers, hedges, or confessional frames. If the\n  draft needs a vulnerable beat, ask for a dated fact and state it flat.\n- Keep the user's voice quirks (lowercase starts, `..` soft pauses, one em\n  dash per ~100 words, one natural triad, their 1-3 intentional emoji).\n- Never promise detector results. If the user asks \"will this pass GPTZero,\"\n  answer honestly: nobody can promise that, and the score on a 150-word\n  caption is noise.\n- Respect the surface: do not turn a carousel's slide text into a single caption\n  or vice versa without flagging it.\n- Never present a caption as publishable without the media reminder (Instagram\n  needs an image or video).\n\n## Instagram-specific tells this skill catches\n\n- A hook that needs the second line to make sense (the \"more\" fold eats it).\n- A caption opening with \"Let's talk about..\", \"Here's the thing..\", or \"POV:\"\n  on something that is not a POV.\n- 20-30 hashtags crammed at the top instead of a 3-5 sized set.\n- Emoji sprinkled through every line (4+ per caption).\n- A bare title on carousel slide 1 instead of a promise + loop.\n- Stacked or hollow rule-of-three lists (\"faster, cheaper, better\"); one\n  natural triple with concrete items is fine.\n- Engagement bait (\"comment YES\", \"tag 3 friends\", \"double tap if\").\n- Staccato stacks and one-word lines for drama (the humanizer fingerprint), or\n  a paragraph where every line reads machine-flat.\n\n## Example\n\nSee `references\u002Fexamples.md` for worked before\u002Fafter rewrites.\n\n## Files\n\n- `SKILL.md` - this file (rewrite scrubber + audit-mode entry)\n- `references\u002Fscrub-rules.md` - V3 regex patterns by tier, density scoring, em dash cap, rhythm rules, forbidden insertions\n- `references\u002Fexamples.md` - worked before\u002Fafter rewrites for captions and slides\n- `references\u002Faudit-checklist.md` - the pre-publish checklist with thresholds\n- `sub-skills\u002Fpost-audit.md` - pre-publish audit workflow (detection-only, no rewrite)\n- `sub-skills\u002Fvoice-profile.md` - build\u002Fupdate the user's Voice & Brand Profile (`--mode profile`)\n- `sub-skills\u002Fillustration.md` - optional Pixfaro image workflow\n\n## Voice profile mode (`--mode profile`)\n\n`ig-humanizer --mode profile` builds or updates the user's Voice & Brand Profile at `..\u002F..\u002Freferences\u002Fvoice-profile.md` from 3-6 of their real Instagram posts pasted in (portable, no token) or, if a read token is set, from pulled activity. Once filled, every writing skill in this bundle drafts in the user's voice automatically. See `sub-skills\u002Fvoice-profile.md`. Triggers: \"build my voice profile\", \"learn my voice\".\n\n## Related skills\n\n- `ig-caption-writer` - generates captions that already pass the humanizer\n- `ig-carousel-planner` - generates carousels that already pass the humanizer\n","# Instagram Humanizer V3\n\nRewrites any caption or carousel slide text to remove the AI tells that human\nreaders notice, and audits a finished caption against the 2026 Instagram\nchecklist. Based on Wikipedia's \"Signs of AI writing\" taxonomy, the 2025-2026\nstylometry literature, our own Instagram caption corpus (n=284, prevalence\nonly), and Instagram-specific patterns (the first-125 fold, the\nlowercase-casual caption register, sized hashtags, sends-and-saves structure).\n**V3 (2026-09):** recalibrated on 2026 evidence. Vocabulary is scored by\ndensity, em dashes are capped instead of banned (and the cap is generous here:\n29% of human captions use one), forced rhythm is now a tell instead of a fix,\nand there is an over-correction guard.\n\n**What this skill does not do:** it does not make text \"pass\" GPTZero,\nPangram, Turnitin or Originality. Those are trained classifiers keyed on the\ninstruction-tuning style signature; prompt-style \"sound like a real person\"\nrewrites are caught 92-95% of the time, and light mechanical rewriting raises\ndetectability. On caption-length text (under 300 words) detector scores are\nnoise. The real value is elsewhere: expert human readers cite vocabulary (53%)\nand sentence structure (36%) as what gives AI text away, and on Instagram a\ncaption that reads as a brand account earns neither the save nor the send.\nThis skill removes what those readers react to.\n\n## What changed in V3\n\nEvidence tier in brackets: [strong] = replicated across 2+ independent\n2025-2026 studies or our own corpus; [vendor] = single platform or vendor\ndataset; [weak] = one study or expert-panel report.\n\n- **Vocabulary moved from a delete-list to density scoring.** The 2023-24 words\n  (delve, tapestry, realm, journey) are decaying as humans avoid them [strong].\n  The durable 2026 markers are common words (significant, crucial, notably,\n  comprehensive, insights, robust, leverage, foster, landscape, nuanced,\n  streamline, elevate) plus grammar: nominalisations and \"-ing\" clause openers\n  at 5.3x the human rate [strong]. AI vocabulary appears in 10% of human\n  captions in our corpus [strong], so one marker in a paragraph is not a\n  verdict. Three is.\n- **Em dash is no longer a tell.** GPT-5.4 emits 1.43 per 1,000 words, below\n  the 3.23 human baseline [strong], and **29% of human Instagram captions\n  contain one** [strong: corpus]. A caption's em dash is never a tell on its\n  own, and a blanket ban over-sterilises captions. New rule: cap at about 1 per\n  100 words (1-2 per caption), replace only the excess with a comma, colon,\n  parentheses or a rewrite. Never a period (a split dash stacks fragments).\n  Zero em dashes across a long caption is its own tell now.\n- **Forced burstiness is the #1 2026 tell, not the fix.** Mechanical\n  long\u002Fshort alternation is a learnable humanizer fingerprint [weak], and\n  \"Short. Punchy. Done.\", \"No X. No Y. Just Z.\", one-word lines for drama and\n  \"The result?\" reveals are the current top reader-cited tells [strong].\n  Captions are mid-length, so Pass 2 is an anti-uniformity guard only: fix a\n  paragraph that reads machine-flat, never manufacture variance.\n- **Rule of three is still a tell, at density.** Tricolon runs at 2x\n  expert-human rate across 2026 frontier models [strong], and 23% of human\n  captions contain one [strong: corpus]. So one natural triple with concrete\n  items stays. Stacked, perfectly parallel or hollow triads and a third triad\n  in a caption get scrubbed.\n- **Fingerprint injection was half wrong.** Named entities and concreteness are\n  supported [strong]; an odd-precision number with a referent in the first 125\n  chars is the strongest hook. Bare numbers are not a discriminator, and\n  inserted hedges and confessions backfire: performed hesitancy is 2x more\n  common in LLM text, and sincerity announcements (\"let me be honest\", \"real\n  talk\", \"POV:\" on something that is not a POV) are a named 2026 tell\n  [strong]. Pass 3 asks for a flat, dated, uncomfortable fact instead.\n- **Over-correction guard.** Humanizer output has its own fingerprint [weak].\n  Pass 4 checks whether Passes 1-3 introduced the very patterns they were meant\n  to remove. Edits are proportional to real problems. When in doubt, leave it.\n\n## When to use\n\n- Before publishing any AI-drafted caption or carousel (rewrite mode)\n- Pre-publish review of a finished caption (audit mode, see `sub-skills\u002Fpost-audit.md`)\n- When a caption feels off and you cannot pinpoint why\n\n## Input\n\nAny text: a caption, the on-image text of a carousel, or a Reel caption.\nOptional: target voice samples (the user's past captions).\n\n## Output\n\n- Rewritten text with AI tells removed\n- A diff showing what changed and why\n- Caption char count (flagging if the hook spills past the first 125 chars)\n- Per-paragraph tell density (markers per paragraph; 3+ triggered a rewrite)\n- Reader-read confidence: \"reads human\", \"mixed\", \"reads AI\" (a reader-tell\n  estimate, not a detector score)\n\n## Modes\n\n```bash\n# Default: scrub AI tells (forensic + strict) and fix Instagram-format issues\nig-humanizer \u003Ctext>\n\n# Forensic only - minimum touch, just kill model leakage\nig-humanizer --mode forensic \u003Ctext>\n\n# Audit - detection-only pass-fail review, no rewrite\n# Runs the 2026 Instagram checklist: first-125 hook, caption length, sized\n# hashtags, emoji limits, CTA quality, media reminder.\n# Returns Blockers + Warnings + suggested fixes. See sub-skills\u002Fpost-audit.md.\nig-humanizer --mode audit \u003Ctext>\n\n# Profile - build\u002Fupdate the user's Voice & Brand Profile. See the section below.\nig-humanizer --mode profile\n```\n\n## The four passes\n\n### Pass 1 - SCRUB (score, then delete or replace)\n\nApply the tiered catalogs in `references\u002Fscrub-rules.md`. The unit of\njudgement is the **paragraph (or slide), not the word**: count markers per\nparagraph, rewrite the paragraph at 3+, leave a single marker alone unless it\nis a reveal bridge, negative parallelism, a sincerity marker, or forensic\nleakage.\n\n- **Forensic** (always on): real model leakage no human types. AI tool markers\n  (oaicite, contentReference, turn0search0), knowledge-cutoff disclaimers (\"As\n  of my last update\"), template blanks ([Your Name]), and em dashes above the\n  cap (more than about 1 per 100 words).\n- **Strict** (default on): what readers react to. The durable 2026 vocabulary\n  set scored by density (significant, crucial, notably, particularly,\n  comprehensive, insights, robust, leverage, foster, landscape, nuanced,\n  streamline, elevate, empower), grammar markers (nominalisations,\n  sentence-opening \"-ing\" clauses), the 2026 model-idiom layer (quietly, \"X\n  matters.\", compound, \"a signal\", \"the work\", \"built different\", \"let that\n  sink in\"), reveal bridges on a single hit (\"The result?\", \"Here's what\",\n  \"Stop X, start Y\", \"plot twist:\"), all forms of negative parallelism,\n  stacked or perfectly parallel triads and any third triad in a caption,\n  phrase cleanups (\"in today's fast-paced world\", \"game-changer\", \"level up\",\n  \"dive in\"), emoji storms, and dead closers (\"what do you think?\", \"double\n  tap if you agree\").\n- **Instagram-format scrubs** (always apply): first-125 hook that stands\n  alone, sized hashtags, emoji limits, carousel slide-1 promise, caption\n  length.\n\n### Pass 2 - RHYTHM (anti-uniformity guard only)\n\nDetectors do not score burstiness, and captions are mid-length (our corpus\nmedian is well under 200 words), so rhythm is neither a reach lever nor a\nfix. What readers notice is the mechanical-uniformity tell (every line the\nsame length, machine-flat) and, worse, the staged variance that\nsecond-generation humanizers add. So Pass 2 has two jobs: fix rhythm only\nwhere it reads machine-flat, and remove manufactured variance everywhere. It\nnever adds variance as a tactic.\n\n- Per paragraph: one genuinely long sentence next to a short one is fine and\n  is what human variance looks like. Two or three mid-length sentences in a\n  row are also fine. Edit only when every sentence in the paragraph runs the\n  same length and reads flat, and then edit one sentence, not the paragraph.\n- Standalone fragments: at most 2 per caption, total. \"every time.\" once is a\n  voice quirk. Three in a caption is a pattern.\n- Banned outright (rewrite as full sentences): \"The X? Y.\" reveals; \"No X. No\n  Y. Just Z.\"; \"All the X. None of the Y.\"; \"Simple. Effective. Easy.\"\n  adjective stacks; one-word lines for drama (\"Still.\" \"Exactly.\"); pseudo-\n  Socratic Q&A (\"Why? Because...\"); \"Short. Punchy. Done.\" staccato runs.\n  Fragment runs are the tell.\n- Layout is not rhythm. One idea per line with blank lines between them is\n  native caption formatting and stays. Fragment-for-drama inside those lines\n  is the tell. Keep the layout, fix the sentences.\n- Carousel slide text is short by design; never chop or pad a slide for\n  rhythm. Never alternate long\u002Fshort\u002Flong\u002Fshort across slides or lines. That\n  seesaw is the humanizer fingerprint.\n\nThe check is \"does any paragraph read machine-flat, and did I add a staccato\npattern\", not a variance number.\n\n### Pass 3 - ADD (human fingerprints)\n\nRequire where the content allows:\n- One odd-precision number WITH a named referent: who, what, when, or what it\n  cost (\"0 to 10k in 4 months, posting 3 times a week\", not \"grew fast\" and not\n  \"10k\"). A bare number is not a fingerprint; the referent carries the signal.\n- One named entity (real person, brand, date, tool)\n- One first-person concrete detail (what you saw, what it cost, what broke)\n- One specific, dated, uncomfortable fact stated flat, with no framing sentence\n  before or after it. Not \"real talk, this one hurt: I lost the client.\" Just\n  \"I lost my biggest client on 14 Feb.\" The fact carries the vulnerability. The\n  frame turns it into performed sincerity, which readers now read as the tell.\n- The lowercase-casual register if the voice calls for it\n\nForbidden as openers or pivots (sincerity announcements, a named 2026 tell):\n\"let me be honest\", \"I'll be real\", \"honestly?\", \"to be direct\", \"the honest\nversion is\", \"real talk\", \"not gonna lie\", \"ngl\", \"can I be vulnerable for a\nsecond\", \"unpopular opinion:\" as a preface to a popular one, \"POV:\" on\nsomething that is not a POV. Also forbidden as insertions: hedges the author\ndid not write (\"perhaps\", \"I might be wrong but\", \"it seems\"). Performed\nhesitancy is 2x more common in LLM text than in expert human text; adding it\nmakes the draft read more AI, not less.\n\nIf the input lacks these, ask the user for a number, name, or moment. Do not\nfabricate.\n\n### Pass 4 - SELF-CHECK (over-correction guard)\n\nHumanizer output has its own fingerprint. Before returning, re-read the result\nonce and answer three questions:\n\n(a) Did Pass 2 create staccato stacks, \"The result?\" reveal bridges, one-word\n    lines, or a long\u002Fshort\u002Flong\u002Fshort seesaw? If yes, merge the fragments back\n    into full sentences.\n(b) Did Pass 3 add a framed confession, a sincerity announcement, or a hedge\n    the author never wrote? If yes, strip the frame and keep only the flat\n    fact, or remove the insertion.\n(c) Did scrubbing flatten the author's voice: uniform tone, no reaction, no\n    concrete detail left, every em dash gone, every triad gone, every emoji\n    gone from a voice that uses them, the lowercase register capitalised? If\n    yes, restore what the author had. Zero em dashes and zero triads in a long\n    caption is a tell in its own right.\n\nIf any answer is yes, dial back rather than scrub harder. Edits must be\nproportional to real problems: a clean caption gets two or three touches, not\na quota. When in doubt whether a pattern is the author or the model, leave it.\n\n## Non-negotiable rules\n\nGlobal voice rules: see root `SKILL.md` Voice rules. Additional skill-specific\nrules (V3):\n\n- **Scrubbing is always in scope.** When asked to humanize, de-AI, finalize, or\n  publish a caption, run at least the forensic + strict passes before it ships.\n  This holds when the user wrote the draft themselves, says they love it as-is,\n  or is in a hurry. Author identity, \"it's already good,\" and time pressure are\n  never reasons to skip the scrub. The forensic + strict pass changes no meaning\n  and takes seconds: run it, then ship. If a constraint truly forbids touching\n  the text, say so explicitly and name every tell left in; the default is to\n  scrub, not to wave it through.\n- **Scrub proportionally.** A pass that finds nothing changes nothing. Do not\n  invent edits to justify the run, and do not report a detector score as the\n  result; report the tells found and fixed.\n- Preserve the user's actual claim and meaning. \"Preserve their voice\" covers\n  voice quirks and what they are claiming, NOT reveal bridges, staccato stacks,\n  or a paragraph with 3+ vocabulary markers. Stripping those is not changing\n  their voice; it is the job.\n- Never introduce facts that were not in the input. If a number is missing, ask.\n- Never introduce sincerity markers, hedges, or confessional frames. If the\n  draft needs a vulnerable beat, ask for a dated fact and state it flat.\n- Keep the user's voice quirks (lowercase starts, `..` soft pauses, one em\n  dash per ~100 words, one natural triad, their 1-3 intentional emoji).\n- Never promise detector results. If the user asks \"will this pass GPTZero,\"\n  answer honestly: nobody can promise that, and the score on a 150-word\n  caption is noise.\n- Respect the surface: do not turn a carousel's slide text into a single caption\n  or vice versa without flagging it.\n- Never present a caption as publishable without the media reminder (Instagram\n  needs an image or video).\n\n## Instagram-specific tells this skill catches\n\n- A hook that needs the second line to make sense (the \"more\" fold eats it).\n- A caption opening with \"Let's talk about..\", \"Here's the thing..\", or \"POV:\"\n  on something that is not a POV.\n- 20-30 hashtags crammed at the top instead of a 3-5 sized set.\n- Emoji sprinkled through every line (4+ per caption).\n- A bare title on carousel slide 1 instead of a promise + loop.\n- Stacked or hollow rule-of-three lists (\"faster, cheaper, better\"); one\n  natural triple with concrete items is fine.\n- Engagement bait (\"comment YES\", \"tag 3 friends\", \"double tap if\").\n- Staccato stacks and one-word lines for drama (the humanizer fingerprint), or\n  a paragraph where every line reads machine-flat.\n\n## Example\n\nSee `references\u002Fexamples.md` for worked before\u002Fafter rewrites.\n\n## Files\n\n- `SKILL.md` - this file (rewrite scrubber + audit-mode entry)\n- `references\u002Fscrub-rules.md` - V3 regex patterns by tier, density scoring, em dash cap, rhythm rules, forbidden insertions\n- `references\u002Fexamples.md` - worked before\u002Fafter rewrites for captions and slides\n- `references\u002Faudit-checklist.md` - the pre-publish checklist with thresholds\n- `sub-skills\u002Fpost-audit.md` - pre-publish audit workflow (detection-only, no rewrite)\n- `sub-skills\u002Fvoice-profile.md` - build\u002Fupdate the user's Voice & Brand Profile (`--mode profile`)\n- `sub-skills\u002Fillustration.md` - optional Pixfaro image workflow\n\n## Voice profile mode (`--mode profile`)\n\n`ig-humanizer --mode profile` builds or updates the user's Voice & Brand Profile at `..\u002F..\u002Freferences\u002Fvoice-profile.md` from 3-6 of their real Instagram posts pasted in (portable, no token) or, if a read token is set, from pulled activity. Once filled, every writing skill in this bundle drafts in the user's voice automatically. See `sub-skills\u002Fvoice-profile.md`. Triggers: \"build my voice profile\", \"learn my voice\".\n\n## Related skills\n\n- `ig-caption-writer` - generates captions that already pass the humanizer\n- `ig-carousel-planner` - generates carousels that already pass the humanizer\n",{"name":5,"description":7},"main","a73152352c0612252131e5cf84e91312a6cf17dc","marketing\u002Finstagram-marketing\u002F.codex-marketplace\u002Finstagram-skills\u002Fskills\u002Fig-humanizer","https:\u002F\u002Fgithub.com\u002FHenrique1601\u002FSkiilsToIAs","https:\u002F\u002Fgithub.com\u002FHenrique1601","https:\u002F\u002Fbezerraportifolio.netlify.app\u002F","Full Stack Developer | React, Next.js, TypeScript, Node.js & IA | Construindo produtos digitais e soluções SaaS","2026-11-15 02:56:04",{"source":8,"skillName":5,"commands":33,"prompt":43,"zipAvailable":44},[34,37,40],{"tool":35,"command":36},"npx","npx skills add Henrique1601\u002FSkiilsToIAs --skill ig-humanizer",{"tool":38,"command":39},"bunx","bunx skills add Henrique1601\u002FSkiilsToIAs --skill ig-humanizer",{"tool":41,"command":42},"pnpm","pnpm dlx skills add Henrique1601\u002FSkiilsToIAs --skill ig-humanizer","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FHenrique1601\u002FSkiilsToIAs\u002Fblob\u002Fmain\u002Fmarketing\u002Finstagram-marketing\u002F.codex-marketplace\u002Finstagram-skills\u002Fskills\u002Fig-humanizer\u002FSKILL.md",true,[46,52,56,60,64,68],{"path":47,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":50,"sourceHtmlUrl":51},"references\u002Faudit-checklist.md","REFERENCE","参考文档",4133,"https:\u002F\u002Fgithub.com\u002FHenrique1601\u002FSkiilsToIAs\u002Fblob\u002Fmain\u002Fmarketing\u002Finstagram-marketing\u002F.codex-marketplace\u002Finstagram-skills\u002Fskills\u002Fig-humanizer\u002Freferences\u002Faudit-checklist.md",{"path":53,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":54,"sourceHtmlUrl":55},"references\u002Fexamples.md",3427,"https:\u002F\u002Fgithub.com\u002FHenrique1601\u002FSkiilsToIAs\u002Fblob\u002Fmain\u002Fmarketing\u002Finstagram-marketing\u002F.codex-marketplace\u002Finstagram-skills\u002Fskills\u002Fig-humanizer\u002Freferences\u002Fexamples.md",{"path":57,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":58,"sourceHtmlUrl":59},"references\u002Fscrub-rules.md",22483,"https:\u002F\u002Fgithub.com\u002FHenrique1601\u002FSkiilsToIAs\u002Fblob\u002Fmain\u002Fmarketing\u002Finstagram-marketing\u002F.codex-marketplace\u002Finstagram-skills\u002Fskills\u002Fig-humanizer\u002Freferences\u002Fscrub-rules.md",{"path":61,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":62,"sourceHtmlUrl":63},"sub-skills\u002Fillustration.md",3797,"https:\u002F\u002Fgithub.com\u002FHenrique1601\u002FSkiilsToIAs\u002Fblob\u002Fmain\u002Fmarketing\u002Finstagram-marketing\u002F.codex-marketplace\u002Finstagram-skills\u002Fskills\u002Fig-humanizer\u002Fsub-skills\u002Fillustration.md",{"path":65,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":66,"sourceHtmlUrl":67},"sub-skills\u002Fpost-audit.md",5445,"https:\u002F\u002Fgithub.com\u002FHenrique1601\u002FSkiilsToIAs\u002Fblob\u002Fmain\u002Fmarketing\u002Finstagram-marketing\u002F.codex-marketplace\u002Finstagram-skills\u002Fskills\u002Fig-humanizer\u002Fsub-skills\u002Fpost-audit.md",{"path":69,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":70,"sourceHtmlUrl":71},"sub-skills\u002Fvoice-profile.md",2346,"https:\u002F\u002Fgithub.com\u002FHenrique1601\u002FSkiilsToIAs\u002Fblob\u002Fmain\u002Fmarketing\u002Finstagram-marketing\u002F.codex-marketplace\u002Finstagram-skills\u002Fskills\u002Fig-humanizer\u002Fsub-skills\u002Fvoice-profile.md",[73,80,87,95,102,109],{"skillNo":74,"name":75,"nameZh":6,"description":76,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":77,"categories":78,"keywords":79,"publishedAt":19},"01M7E2ANBW9V6AVX5JXNCEAWF8","rag-architect-agent","Agente Arquiteto de RAG, Embeddings & Busca Vetorial. Especialista em design e implementação de pipelines de Retrieval-Augmented Generation de alta fidelidade: chunking semântico, modelagem pgvector no PostgreSQL (Neon\u002FSupabase), busca híbrida (cosseno + BM25), reranking, validação via TypeSafe AI e filosofia Ponytail (anti-overengineering).","https:\u002F\u002Fgithub.com\u002FHenrique1601\u002FSkiilsToIAs\u002Fblob\u002Fmain\u002Fcustom-agents\u002Frag-architect-agent\u002FSKILL.md",[],[],{"skillNo":81,"name":82,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"publishedAt":19},"01M7E29KKF8R02ZRGYE2MJM67G","rag-architecture-patterns","Padrões arquiteturais e implementação de pipelines RAG (Retrieval-Augmented Generation) de alta fidelidade para qualquer projeto ou sistema. Cobre chunking semântico, embeddings, busca híbrida (pgvector + BM25), reranking, validação via TypeSafe AI e filosofia Ponytail (anti-overengineering).","https:\u002F\u002Fgithub.com\u002FHenrique1601\u002FSkiilsToIAs\u002Fblob\u002Fmain\u002Fai-agents\u002Frag-architecture-patterns\u002FSKILL.md",[],[],{"skillNo":88,"name":89,"nameZh":6,"description":90,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":91,"licenseFileUrl":6,"sourceHtmlUrl":92,"categories":93,"keywords":94,"publishedAt":19},"01M7CWRA8E3BJ7CPY8N8EG28J1","skill-repair","Use this to fix and re-install agent skills that have failed installation.\nThis skill provides the necessary context and permissions to surgically update\nthe `manifest.json` after a fix has been applied.","Apache-2.0","https:\u002F\u002Fgithub.com\u002FHenrique1601\u002FSkiilsToIAs\u002Fblob\u002Fmain\u002Futilities\u002Fskill-repair\u002FSKILL.md",[],[],{"skillNo":96,"name":97,"nameZh":6,"description":98,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":99,"categories":100,"keywords":101,"publishedAt":19},"01M7CWR66FZP8FA47WVZ4Z2EXC","security-scan","Scan your Claude Code configuration (.claude\u002F directory) for security vulnerabilities, misconfigurations, and injection risks using AgentShield. Checks CLAUDE.md, settings.json, MCP servers, hooks, and agent definitions.","https:\u002F\u002Fgithub.com\u002FHenrique1601\u002FSkiilsToIAs\u002Fblob\u002Fmain\u002Futilities\u002Fsecurity-scan\u002FSKILL.md",[],[],{"skillNo":103,"name":104,"nameZh":6,"description":105,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":106,"categories":107,"keywords":108,"publishedAt":19},"01M7CWR0QPY58QPTR1P759TFNG","security-review","Use this skill when adding authentication, handling user input, working with secrets, creating API endpoints, or implementing payment\u002Fsensitive features. Provides comprehensive security checklist and patterns.","https:\u002F\u002Fgithub.com\u002FHenrique1601\u002FSkiilsToIAs\u002Fblob\u002Fmain\u002Futilities\u002Fsecurity-review\u002FSKILL.md",[],[],{"skillNo":110,"name":111,"nameZh":6,"description":112,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":113,"categories":114,"keywords":115,"publishedAt":19},"01M7CWQXCE40XWT4FZ3VWZFGDP","security-bounty-hunter","Hunt for exploitable, bounty-worthy security issues in repositories. Focuses on remotely reachable vulnerabilities that qualify for real reports instead of noisy local-only findings.","https:\u002F\u002Fgithub.com\u002FHenrique1601\u002FSkiilsToIAs\u002Fblob\u002Fmain\u002Futilities\u002Fsecurity-bounty-hunter\u002FSKILL.md",[],[]]