humanize-copy
Skill12 starsChanged 11 days agoMITView on GitHub
Humanize prose by removing LLM tells. Use for landing copy, docs, READMEs, emails, UI microcopy, release notes, social posts, or requests like humanize, less robotic, remove AI tells, punch up, or make this sound human.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- millwright-labs/humanize-copy-skilltheir own version, first seen Aug 24, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- millwright-labs/humanize-copy-skill★ 0same name
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 repository21 other files it gives its agents.
- AGENTS.mdAGENTS.md
- llms.txtllms.txt
- a11y-passSkill
- art-directionSkill
- cognitive-a11ySkill
- color-systemSkill
- component-statesSkill
- data-vizSkill
- design-system-interviewSkill
- deslop-uiSkill
- empty-statesSkill
- form-uxSkill
- i18n-readySkill
- improve-existing-websiteSkill
- micro-motionSkill
- responsive-layoutSkill
- spacing-systemSkill
- tastecheck-passSkill
- tasterollSkill
- themingSkill
- web-typographySkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 16
Do not synonym-swap detached text.
- Distinctive: line 69
Read-aloud test, direct opening, and no formulaic cadence, hedge, or thesaurus-salad?
- Distinctive: line 91
Run the three-pass audit and record rule IDs.
- Very specific: line 101
Stop / handoff: Stop when source facts or approval constraints are missing.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Humanize Copy
Human copy sounds meant for this reader, now. Keep facts; replace formula with consequence, stance, and cadence the speaker can sustain aloud.
Establish the speaking situation before editing
Do not synonym-swap detached text. Build a voice brief from source, reader job, protected language, defensible claim, and register. For a real author, use supplied/approved samples; never infer private identity from stereotypes.
| Voice brief field | Evidence to use |
| Speaker and audience | Named role, relationship, and reader context from the request/source |
| Reader job | Decision, action, reassurance, or understanding the text must support |
| Defensible claim | Source fact, direct quote, measurement, or explicitly marked opinion |
| Protected language | Legal, safety, product, customer, or brand wording that cannot be casually altered |
| Register boundary | Degree of formality, humor, intimacy, and certainty the situation can bear |
Without facts/voice authority, change only clarity, order, and cadence; never invent claims, metrics, anecdotes, testimony, or personal voice.
Edit in three passes
- Source: mark protected meaning and authority behind every claim.
- Stance: decide what the speaker can say plainly; lead with reader consequence, decision, or next action.
- Shape: cut wrapper, restore causal order, tune sentence length/register.
Compare revision to source; apply references/writing-constitution.md, then use kill-list and
rhythm references as detectors after meaning is settled.
Detect formula before replacing words
The kill list is a detector, not a ban. Inspect buzzwords, ceremonial hooks, manufactured contrast, repeated rule-of-three/em-dash cadence, hedges, and generic metaphor. Preserve technical, legal, and brand terms. When it fires, name the subject, state the consequence, or remove the sentence; a fancier synonym is not repair.
Worked transformation
Source facts: appointments are booked online; technicians give a two-hour arrival window; customers receive a text before arrival.
Before: “Experience seamless repair solutions designed to get your day back on track.”
After: “Book online. We’ll give you a two-hour window and text before the technician arrives.”
The revision earns confidence from source: it replaces unsupported emotional promise with decision, timing, and useful reassurance. Target specificity and rhythm that fit authority, not universally casual copy.
Decision order and evidence
Build the brief/fact ledger; revise stance, order, precision, and cadence without changing facts or safety meaning. Record source authority for new specificity.
Self-check
- Read-aloud test, direct opening, and no formulaic cadence, hedge, or thesaurus-salad?
- Source-backed specifics and bounded claims—not competitor-generic copy or invented texture?
- Register fits this reader, task, and consequence?
- Writing rules applied and their IDs recorded in the evidence?
Reference files
references/kill-list.md— detector list and examples.references/rhythm-and-voice.md— structure, stance, cadence.references/decision-records.md— novel-case ADR rules.references/writing-constitution.md— portable IA floor.
How to deliver
Deliver revised copy first when requested. Follow with voice brief and 2–5 high-leverage
mechanism changes when useful. Preserve facts/register; pair non-English work with
i18n-ready.
Information-architecture gate
Apply the bundled Writing Constitution floor (references/writing-constitution.md):
old-before-new, stress position, anchor/linking/umbrella/preview, and exact key-term repetition.
Run the three-pass audit and record rule IDs. This reference is portable; do not use a
workstation-local copy. Echo argument terms; remove formulaic repetition.
<!-- contract:v1:start -->Contract (generated)
Canonical detail: contract.json.
- Route: UI or product copy is factually correct but generic, hedged, clichéd, or unlike the intended author voice.; avoid: The request requires changing product facts, legal meaning, or interaction behavior.
- Exclude: Do not invent claims, testimonials, or product facts. (+2 in contract.json)
- Stop / handoff: Stop when source facts or approval constraints are missing. (+1 in contract.json); receives [cognitive-a11y, design-system-interview, deslop-ui, empty-states, improve-existing-website, tasteroll] -> sends [form-ux, i18n-ready, tastecheck-pass]
- Output: fact-preserving copy revision with voice rationale
- Evidence:
table_with_evidencewithstatus,reason,remediation,evidence,provenance.
<!-- contract:v1:end -->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.

