Humanize ai writing
Skill hewieextracellular787/humanize-ai-writing/humanize-ai-writing
Force any writing to follow human patterns and strip every known sign of AI-generated text ("AI slop"). Use whenever producing or editing prose that will be read by a person, articles, blog posts, emails, captions, essays, documentation, social copy, reports, and whenever the user says "humanize this", "make it sound human", "remove AI slop", "doesn't sound like AI", "sounds robotic", or "rewrite this naturally". Apply this as a mandatory final pass on any long-form prose before delivering it.From its SKILL.md
npx -y skills add hewieextracellular787/humanize-ai-writing --skill humanize-ai-writingAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
5.1 KB, ~1.2k tokens by cl100k_base, as published. Nobody here has run it
Humanize AI Writing
Make every piece of prose read like a thoughtful person wrote it. This skill is a guardrail while drafting and a rewrite pass on existing text. It encodes the AI tells cataloged in Wikipedia's "Signs of AI writing" as rules you must follow.
When this triggers
- The user asks to humanize, de-slop, or de-robotify text.
- The user says writing "sounds like AI" or "sounds generic."
- You're about to deliver any long-form prose. Run the final pass before sending.
Workflow
- Read the target text (the user's, or your own draft).
- Detect: scan against
references/ai-tells.md. Mark each hit. - Rewrite: apply
references/rewrite-rules.md. Fix tells sentence by sentence. Keep meaning, facts, numbers, names, and the author's position. - Self-check: run
assets/checklist.md. Tells cluster, so scan twice. - Deliver clean text. If you cut a claim for lacking a source, say so in one line. Never describe this process inside the prose.
THE RULES, follow all of them
- Banned vocabulary: delve, tapestry, testament, underscore, pivotal, intricate, meticulous, garner, vibrant, bolstered, enduring, interplay, crucial, valuable, exemplify, renowned, groundbreaking, boasts, nestled, landscape (figurative), realm, embark, leverage (verb), elevate, unlock, unleash, harness, robust, seamless, holistic, myriad, plethora, cornerstone, beacon, paradigm, synergy. Editorializing openers: additionally, moreover, furthermore, notably, importantly. Canned phrases: "in today's world", "in the heart of", "rich cultural heritage", "when it comes to", "it's important to note", "it's worth noting".
- No fake significance / legacy: no "stands as", "serves as a testament", "marks a pivotal moment", "plays a vital role", "leaves an indelible mark", "setting the stage for", "turning point", "evolving landscape", "focal point", "deeply rooted", "reflects a broader", "represents a shift".
- No present-participle padding: no trailing "-ing" clauses with no fact ("…, highlighting its importance", "…, reflecting a commitment to…").
- No negative parallelism: no "not just X, but Y", "not only… but also", "it's not X, it's Y", "no X, no Y, just Z", "more than just", "X rather than Y".
- No rule-of-three filler: use the count the meaning needs, not always three.
- Punctuation & typography: avoid em dashes (rewrite with comma / period / colon / parentheses); straight quotes and apostrophes only; almost no bold; sentence-case headings; no emoji as formatting; no inline-header bold lists.
- No promotional / puffery tone: no "commitment to", "cutting-edge", "world-class", "state-of-the-art", "passionate about", "we're excited to", "revolutionary", "diverse array", "seamlessly connecting".
- Use plain copulas: "is/are/has", not "serves as", "stands as", "represents", "features", "maintains", "offers", "boasts", "refers to".
- Specifics over hedging, no notability padding: replace "experts say / studies show / observers have cited / some critics argue / industry reports" with a named source or cut it. No lists of media the subject appeared in. Never write "maintains an active social media presence". Never invent a source.
- No hollow conclusions or formula sections: no "In conclusion", "Overall", "Ultimately", "In summary"; no "Despite its [positives]… faces challenges…" → "Future Outlook" template; no future-speculation unless asked.
- Vary rhythm: mix sentence lengths, use short ones, allow fragments. Let a word repeat instead of cycling synonyms (no elegant variation).
- No artifacts: strip contentReference, oaicite, oai_citation, turn0search0, utm_source= in links, markdown leaking where it shouldn't, curly-quote noise.
Voice defaults
- Active voice, concrete subjects. Contractions are fine.
- Don't restate the topic at the start of every section.
- Don't add hype to fix flatness or fake slang to "sound human". Plain is the goal.
Important, don't overcorrect
A single flagged word is not proof of AI, and detectors are unreliable. Avoid clusters of tells while keeping meaning, facts, and the author's position. Don't mangle a sentence or delete real information just to purge one banned word rewrite cleanly, or leave it if the rewrite would lose meaning.
Final pass (do not skip)
Re-scan and rewrite any hit for rules 1–12 before sending. Never mention this pass.
Reference files
references/ai-tells.md, full detection catalog.references/rewrite-rules.md, fix + before/after for each tell.assets/checklist.md, the pre-ship checklist.
What ships with it: 3 files
11.1 KB alongside SKILL.md
assets/
- checklist.md2.4 KB
references/
- ai-tells.md5.2 KB
- rewrite-rules.md3.6 KB
Gives 0 of the 12 instructions most docs writing skills give in ~1.2k tokens
Counted across 1,637 of the 3,044 authors here whose files we hold, read 2026-08-07
- Announce the skill at startin 54 of 1637, across 26 files
- Convert legacy doc files before editingin 45 of 1637, across 7 files
- Predict questions readers might askin 42 of 1637, across 4 files
- Generate clarifying questions for initial contextin 42 of 1637, across 3 files
- Create document scaffold with placeholder textin 42 of 1637, across 3 files
- Brainstorm content options for each sectionin 42 of 1637, across 3 files
- Test the document with a fresh context-less instancein 42 of 1637, across 3 files
- Include exact file paths in every taskin 42 of 1637, across 15 files
- Ask interview questions one at a timein 42 of 1637, across 27 files
- Apply surgical edits during refinementin 41 of 1637, across 2 files
- Offer structured workflow or freeformin 40 of 1637, across 1 file
- Ask for document meta-contextin 40 of 1637, across 2 files
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.