humanize-writing
Skill36 starsChanged 6 months agoMITView on GitHub
Revise AI-generated or mechanical-sounding text so it reads like a specific person wrote it. Use this skill when the user asks to "humanize", "make natural", "remove AI feel", "rewrite like a person wrote it", fix "AI-sounding" or "기계적인" text, or pass AI detection. Also trigger when the user pastes text and asks for a more natural or human tone, in either English or Korean. Do NOT use for summarization, translation, or creative writing from scratch.
18 other projects use a version of this.
Adopted by
Most recently taken up by
- aktsmm/Agent-Skillstheir own version, first seen Nov 8, 2026
- east18047-zmjjkk/humanize-writingtheir own version, first seen Nov 8, 2026
- huangzehao/skillstheir own version, first seen Nov 8, 2026
- Littlpinguin/marketing-cockpit-templatetheir own version, first seen Nov 3, 2026
- Yeseh/skillstheir own version, first seen Nov 1, 2026
- Tejas9406/vibe-coding-ostheir own version, first seen Nov 1, 2026
Other versions18 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 18
- jpeggdev/humanize-writing★ 62same name
- lguz/humanize-writing-skill★ 52same name
- aktsmm/Agent-Skills★ 26same name
- shir-danishyar/humanize★ 14same name
- aws-solutions/konductor★ 11same name
- Littlpinguin/marketing-cockpit-template★ 9same name
- geyadawang-boop/humanize-writing★ 7same name
- agentbody/skills★ 1same name
- east18047-zmjjkk/humanize-writing★ 1same name
- huangzehao/skills★ 1same name
8 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.
In this repository29 other files it gives its agents.
- skills/hpc-monitor/AGENTS.mdAGENTS.md
- aseSkill
- blender-atom-renderSkill
- catagentSkill
- ccel_slide_skillSkill
- computational-section-writingSkill
- design2htmlSkill
- docx-scientific-formattingSkill
- figure_constructionSkill
- hpc-monitorSkill
- html-minimalSkill
- humanizer_engSkill
- ko-writingSkill
- rxn-diagram-insetSkill
- matplotlib-scientificSkill
- meetingnote-paperworkSkill
- openrouter-costSkill
- ovito-atom-renderSkill
- paper-bibSkill
- paperSkill
- pptx-too-heavySkill
- report_less_slopSkill
- reportSkill
- slide-auditSkill
- smart-compactSkill
- soagentSkill
- transcript2htmlSkill
- youtube2mp4Skill
- youtubeSkill
The file
We split this skill into the rules it gives an agent
8 rules. 0 are common, 8 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 11 edited versions we read keep these.
8
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
Don't fabricate: Never invent facts, quotes, sources, personal experience, numbers, named actors, or emotions that aren't in the original.
- Distinctive: line 33
If more than half the paragraphs need rebuilding, tell the user the text needs rewriting, not polishing, and confirm before proceeding.
- Distinctive: line 37
Opening: If the first sentence could open any piece on the same general topic, replace it with the first specific claim, finding, event, or …
- Distinctive: line 39
Second paragraph: Must advance the opening, not retreat to context or methodology.
- Distinctive: line 45
Counter-arguments: If the text dismisses a counter-argument with "both perspectives have merit" or equivalent, either engage it honestly …
- Distinctive: line 49
Sources: If a source is cited, it needs author, work, year, and specific claim.
- Distinctive: line 64
Add a brief note on what changed only if the user asked for explanation.
- Distinctive: line 65
Do not describe your process or mention this skill.
“Common” is measured: kept by at least 6 in 10 of the 11 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.
Files
Human Writing Revision
Revise text so it reads as if a specific person wrote it for a specific reader.
Three Rules (in priority order)
- Don't fabricate. Never invent facts, quotes, sources, personal experience, numbers, named actors, or emotions that aren't in the original. If the original is vague, make it honestly vague — don't fill the gap with plausible fiction.
- Remove mechanical signals. Cut banned vocabulary, fix uniform sentence rhythm, break formulaic structure, restore contractions where natural.
- Restore specificity. Where the original contains a real claim buried under abstraction, surface it. Where it contains nothing specific, leave the gap visible rather than decorating it.
Process
Step 1: Identify audience and edit depth
Infer the audience from three signals in the text:
- Terms used without definition → assumed reader knowledge
- Consequences stated without explanation → assumed reader stakes
- Pronoun pattern (you/we/one) → assumed relationship
Then choose the lightest sufficient edit:
| Signal | Action |
| Structure sound, surface awkward | Light polish — vocabulary, rhythm, contractions |
| Multiple paragraphs with no actor, event, or consequence | Rewrite — rebuild those paragraphs from their intended claim |
| User asks "what's wrong with this?" | Diagnostic — explain problems, don't rewrite unless asked |
If more than half the paragraphs need rebuilding, tell the user the text needs rewriting, not polishing, and confirm before proceeding.
Step 2: Edit
Opening. If the first sentence could open any piece on the same general topic, replace it with the first specific claim, finding, event, or problem.
Second paragraph. Must advance the opening, not retreat to context or methodology. Context goes third or later.
Sentences. Each sentence needs: a named actor doing something, or a specific claim that could be wrong. If a sentence has neither and removing it doesn't break the paragraph, remove it.
Rhythm. Vary sentence length to reflect emphasis. Short after long lands harder. Don't vary for its own sake.
Counter-arguments. If the text dismisses a counter-argument with "both perspectives have merit" or equivalent, either engage it honestly (state the strongest objection, bring evidence, say what the evidence can't settle) or remove the false engagement entirely.
Endings. End when the point has landed. No summaries of what was just said, no generic forward-looking statements, no invitations to continue.
Sources. If a source is cited, it needs author, work, year, and specific claim. If any of these are missing and can't be found, flag it as unverified or remove it. Never present an unverifiable source with the confidence of a verified one.
Step 3: Check for over-correction
Read the result and ask: is every specific detail earned by the original text, or was it inserted because this skill said to be specific? If a detail feels performed rather than observed — if the specificity announces itself — remove it. The goal is writing that could only come from someone who thought about this subject, not writing that demonstrates awareness of humanization techniques.
Language-Specific References
- For Korean text: read
references/korean-patterns.mdbefore editing. - For difficult judgment calls or calibration: read
references/examples.md. - For deeper principles behind any rule above: read
references/guide.md.
Output
Return the revised text first, without preamble. Add a brief note on what changed only if the user asked for explanation. Do not describe your process or mention this skill.
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
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.

