agentleFS
pivoshenkopivoshenko/pivoshenko.ai/skills/humanize/SKILL.md

humanize

Skill6 starsChanged 44 days agoMITView on GitHub

Remove AI-writing tells from any prose, normalize punctuation to plain ASCII (em/en dash -> plain punctuation per the replacement order, curly quotes -> straight, ellipsis char -> "..."), and raise headings to Title Case. Use when the user says "humanize this", "make it sound human", "de-AI this", "this reads like ChatGPT", "remove AI patterns", or when editing/reviewing any prose (doc, README, PR description, commit message, email, post) that shows AI tells. The general-purpose de-AI pass for any prose, in any format, unless a more specific writing workflow already owns the piece. Skips agent-instruction files (CLAUDE.md, AGENTS.md, context.md, llms.txt, cursor rules) unless the user aims it at one - their terse fragment style is the working format, not a tell.

We split this skill into the rules it gives an agent

7 rules. 0 are common, 5 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 63 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Preserve information, not shape: Every claim survives. Compress dull parts, merge/split paragraphs freely.

    line 14
  2. Distinctive:

    Never invent facts: No new names, numbers, dates, quotes, citations.

    line 15
  3. Very specific:

    Before delivering, scan the result for — – …, curly quotes, →.

    line 41
  4. Distinctive:

    Every heading and title is Title Case.

    line 45
  5. Distinctive:

    Keep lowercase mid-title: articles (a, an, the), coordinating conjunctions (and, or, but, nor, for, so, yet), prepositions of four letters …

    line 48
  6. Very specific:

    Never recase: names that are lowercase by design (npm, ripgrep, fish), code identifiers, acronyms, and titles of works quoted from the …

    line 49
  7. Distinctive:

    No hits at all -> say the text is clean, skip the table

    line 119

“Common” is measured: kept by at least 6 in 10 of the 63 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Humanize

Strip AI tells from text. Keep every fact, change the prose.

Ground Rules

  1. Preserve information, not shape. Every claim survives. Compress dull parts, merge/split paragraphs freely. Information beats structure
  2. Never invent facts. No new names, numbers, dates, quotes, citations. Sentence needs missing detail to work -> ask, or write the plain version without it
  3. Match voice. Formal/casual/technical as intended. User provides own writing sample -> its habits outrank every rule below, including the dash rule
  4. Don't sterilize. Voiceless is as obvious as slop. Essay/opinion/post -> keep opinions, asides, uneven rhythm. Technical/reference -> plain and neutral IS the human voice there

Out of Scope: Agent Instruction Files

Files written for a model to execute, not for a person to read: CLAUDE.md, AGENTS.md, context.md, llms.txt, .cursor/rules/*, GEMINI.md, .github/copilot-instructions.md, and any other agent/system-prompt file. Terse fragments, arrows, abbreviations, and lowercase headings are the working format there - compression is the point, and recasing headings or padding fragments into sentences costs tokens and blurs directives. Skip these files entirely: no rewrite, no heading recase, no punctuation pass.

Same for the prose inside a normal file when it is addressed to a model: system prompts, skill bodies, prompt templates in code or fixtures.

Exception - the user aims the skill at one of these deliberately ("humanize my CLAUDE.md", "clean up this system prompt"). Do it, and say once that the file is agent-facing so terseness there was probably intentional. A doc that merely mentions or quotes such a file is ordinary prose; edit it normally.

House Conventions

The two rules below are house style, not AI tells. They run on every pass, on every file this skill touches, whether or not the prose shows a single AI pattern - which is exactly why they need saying out loud: someone who asked for a de-slop pass and got their headings recased should have seen it coming from this section.

Both are defaults, not laws. "punctuation only", "leave the casing", "don't touch my headings" -> honor it for that pass and note in the findings table which convention was skipped. A file with a documented style guide of its own outranks both (see the exceptions under each).

Punctuation -> Plain ASCII

  • Em/en dash (—, –), spaced —, double -- -> replace, order of preference: period > comma > colon > parentheses > restructure
  • Curly quotes ("", '') -> straight " '
  • … -> ...
  • → -> ->
  • Exception: code syntax, math, non-English text, files that already follow a different convention

Before delivering, scan the result for — – …, curly quotes, →. Any hit -> not done.

Headings -> Title Case

Every heading and title is Title Case; sentence-case headings get raised.

  • Capitalize the first word, the last word, and every major word (nouns, verbs, adjectives, adverbs, pronouns, subordinating conjunctions)
  • Keep lowercase mid-title: articles (a, an, the), coordinating conjunctions (and, or, but, nor, for, so, yet), prepositions of four letters or fewer (of, in, to, on, at, by, from, with)
  • Never recase: names that are lowercase by design (npm, ripgrep, fish), code identifiers, acronyms, and titles of works quoted from the source - such a name stays lowercase even in first position
  • Scope: markdown headings, document/section titles, table-of-contents entries. Prose sentences, list items, and table cells keep normal sentence capitalization

Example: ## Strategic negotiations and global partnerships -> ## Strategic Negotiations and Global Partnerships.

Exception - a surface with a documented style guide that calls for a different casing keeps that casing, whatever it is: sentence case, all-lowercase, all-caps section labels. Brand and product copy, UI labels, and house-styled publications usually have one, and there the heading style is part of what readers recognize - an all-lowercase house voice reads as deliberate, and raising it is the single most visible way this pass can damage a piece it was meant to help. A stated convention outranks this default; a file that merely happens to use a casing does not - that file is the reason the rule exists. Unsure whether a convention is stated or accidental -> leave the casing alone and say so in the findings table; a heading left unraised costs nothing, a house voice flattened costs the author.

Tells to Kill

Full before/after catalog: references/patterns.md. Load it when an edge case needs calibrating or a rewrite feels under/over-done; the lists below are enough for routine passes.

Content:

  • Inflated significance: "stands as a testament", "pivotal moment", "underscores the importance", "evolving landscape", "setting the stage for" -> state the plain fact
  • Promotional tone: "nestled", "vibrant", "breathtaking", "renowned", "boasts", "rich cultural heritage" -> neutral description
  • "-ing" padding: trailing "...showcasing/highlighting/reflecting/ensuring..." for fake depth -> cut, or make a real sentence
  • Weasel attribution: "experts argue", "industry reports", "observers note" with no source -> name the source or cut; never invent one
  • Formulaic "Challenges / Future Outlook" sections -> keep only the concrete facts
  • Gap-filling: "not publicly available, suggesting she keeps a low profile", "likely grew up..." -> say what isn't known or cut; never dress a guess as fact

Vocabulary:

  • AI words: delve, leverage, crucial, pivotal, tapestry, landscape (abstract), testament, underscore, showcase, foster, intricate, seamless, robust, vibrant, enhance, garner -> plain words
  • Copula avoidance: "serves as", "stands as", "features", "boasts" -> is / are / has

Structure:

  • "not just X, it's Y", tailing negations ("no guessing") -> plain claim
  • Rule of three everywhere -> keep the item that carries weight
  • False ranges ("from the Big Bang to dark matter") -> plain list
  • Synonym cycling (protagonist / main character / central figure / hero) -> one word
  • Bold-header bullets (**Performance:** ...) -> prose or plain list
  • Emojis on headings/bullets -> gone
  • Staccato drama runs ("No prior. No nostalgia. Gone.") -> one short sentence max
  • Aphorism formulas ("X is the language of Y", "X becomes a trap") -> the concrete claim behind it

Filler:

  • "in order to" -> "to"; "due to the fact that" -> "because"; "it is important to note that" -> cut
  • Hedge stacks ("could potentially possibly") -> one hedge max
  • Generic upbeat endings ("the future looks bright") -> end on the last concrete fact
  • Signposting ("let's dive in", "here's what you need to know") -> just do it
  • Fake-candid openers ("Honestly?", "Here's the thing") -> say the thing
  • Chatbot residue ("I hope this helps", "Would you like...", "Great question!") -> gone

Don't Over-Flag

One tell means nothing - look for clusters. A single em dash, one triad, polish, formal vocabulary, one short punchy sentence = normal human writing. Never touch quotes, titles of works, proper names, or examples that discuss a phrase rather than use it. Preserve human signals: weird specific detail, mixed feelings, asides, self-corrections, varied sentence length.

Modes

  • Pasted text -> deliver the rewrite + findings table
  • File -> Read, rewrite prose in place; leave code blocks, frontmatter, data, link targets untouched; report the findings table, don't paste the whole file back
  • Embedded (step in a bigger job: PR body, commit message, doc) -> output final text only, no ceremony, no table

Findings Table

Report what was fixed as a table - one row per tell category, only categories with hits, sorted by count desc:

TellCountExamples
em/en dashes12"policy—announced" -> "policy, announced"
AI words5delve x2, leverage x2, robust
headings recased4"Getting started" -> "Getting Started"
rule of three2"innovation, inspiration, and insights"
filler3"in order to" x3
  • Count = instances fixed, not sentences touched
  • Punctuation rows (— – … curly quotes, -> arrows): get exact counts by grepping the source before rewriting, don't eyeball
  • Examples: 1-3 short samples per row, enough to spot-check; truncate long ones
  • No hits at all -> say the text is clean, skip the table

Process

  1. Scan for tells above
  2. Draft rewrite - reads naturally aloud, varied sentence length, is/are/has
  3. Audit: anything still AI-sounding? any fact not in the source? any — – … left? Fabrication = defect even when it sounds better
  4. Final rewrite

Download now (ZIP)

  • Other versions72 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 72

    1. epoko77-ai/im-not-ai★ 6ksame name
    2. pedrohcgs/claude-code-my-workflow★ 1.6ksame name
    3. PolyArch/humanize★ 1.5ksame name
    4. cwinvestments/memstack★ 425same name
    5. Mathews-Tom/armory★ 340same name
    6. whawkinsiv/solo-founder-skills★ 260same name
    7. whawkinsiv/claude-code-skills★ 258same name
    8. c-kick/hnl-agent-skills★ 56same name
    9. tzengyuxio/skills★ 45same name
    10. amondnet/yoonmoon★ 17same name

    62 more not listed

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • Used alongsideProjects with it also use humanize-redo, humanize-korean and 10 more.

    Other files people use alongside this one.

    Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
    • dawn840705Skill

      humanize-redo

      가장 최근 윤문 결과를 2차로 다시 다듬는다 — 특정 카테고리·문단·강도 조정도 가능. humanize-korean 의 윤문 콜을 기존 run_id 에 재실행해 잔존 finding 을 처리한다. 트리거 — "$humanize-redo". **명시적 슬래시 호출 전용** — 자연어 후속 요청은 `humanize-korean` 오케스트레이터가 받는다.

      Also in 5 projects that use this

    • mskim8717Skill

      humanize-korean

      AI(ChatGPT·Claude·Gemini 등)가 쓴 한글 텍스트를 "사람이 쓴 글처럼" 윤문해주는 오케스트레이터 스킬. 번역투·영어 인용 과다·기계적 병렬·관용구·피동태 남용·접속사 남발·리듬 균일성·이모지/불릿 과다 등 10대 카테고리 40+ AI 티 패턴을 탐지·분류해 내용은 한 글자도 건드리지 않고 문체·리듬·표현만 자연스러운 한국어로 재작성한다. 트리거 — "AI 티 없애줘", "AI 같은 글 자연스럽게", "GPT/ChatGPT 문체", "AI 번역투 고쳐", "사람이 쓴 것처럼 윤문", "AI 윤문", "ChatGPT 티 제거", "한글 AI 탐지·윤문", "AI 글 사람처럼", "번역투 제거", "영어 인용 많은 글 윤문", "AI 글 티 안 나게", "휴머나이저", "humanize Korean", "AI detector bypass 한글". 후속 작업 — "특정 카테고리만 다시", "윤문 강도 조정", "장르 바꿔서", "이 문단만", "2차 윤문" 도 모두 이 스킬. 단순 맞춤법·오탈자 교정은 직접 처리, 번역은 번역 스킬, 내용 추가·삭제를 동반한 재작성은 별도 집필 스킬.

      Also in 3 projects that use this

    • CoralGameSkill

      perf-profile

      Structured performance profiling workflow. Identifies bottlenecks, measures against budgets, and generates optimization recommendations with priority rankings.

      Also in 3 projects that use this

    • Rayan55050Skill

      content

      Создание контента: посты, статьи, тексты

      Also in 3 projects that use this

    • luischanciSkill

      proofread

      Run the proofreading protocol on manuscript files. Checks grammar, typos, overflow, consistency, and academic writing quality. Produces a report without editing files.

      Also in 3 projects that use this

    • slavingiaSkill

      pricing

      Help figure out pricing for a product or service using minimalist entrepreneur principles. Use when someone is setting prices, considering price changes, or struggling with what to charge.

      Also in 3 projects that use this

  • In this repository21 other files it gives its agents.

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.