How much does Humanizer cost?

Humanizer is free and MIT-licensed, published as a plain Markdown skill on GitHub. There is no service, account or paid tier attached to it. The cost you do pay is whatever your agent charges for the rewrite itself, since Humanizer runs inside Claude Code, Claude Desktop or another harness that installs skills through the Skills CLI.

Does Humanizer help text get past AI detectors?

Humanizer v3.0.0 explicitly removed the "ai-detection" keyword from its package files, and the project frames the skill as a writing-quality tool rather than a detector-evasion tool. The stated goal is text that reads like a person wrote it, and the README is clear that Humanizer does not invent content while rewriting.

Will Humanizer change the facts in my draft?

No. Humanizer rewrites for voice and structure while preserving the claims in the original text, and the README states plainly that the skill "does not make things up". Patterns it removes are stylistic — staged rhetoric, forced triads, decorative bold, chatbot wrappers — rather than the substance of what you wrote.

What changed in Humanizer v3.0.0 compared with version 2?

Humanizer v3.0.0 is a restructure rather than a feature release: 35 patterns became 25 across five ranked sections, the five-part workflow collapsed into one, and duplicated guidance was merged. The release folds in contributor fixes from eight pull requests, closes three issues, drops two rules that turned out to describe human habits, and reorders the README to put installation first.

Can Humanizer match my own writing voice?

Yes. The README describes giving Humanizer two or three paragraphs of your own writing before the text you want rewritten, so the output follows your style rather than a generic plain register. Without samples, Humanizer still rewrites the draft, but it targets the 25 listed patterns instead of imitating a specific author.

blader · 2026-09-06 · major

Humanizer v3.0.0 — the AI-writing cleanup skill drops 35 patterns to 25

Humanizer v3.0.0 rewrites AI-sounding text as an agent skill for Claude Code and Claude Desktop. The release merges 35 writing patterns into 25 without dropping any checks, and adds a validator that caps the skill file at 400 lines.

GitHub social card for blader/humanizer, an agent skill that rewrites AI-sounding text

An agent skill that strips the 25 habits which make text read as AI-written, without changing the facts.

Key specs

GitHub stars	44.7k
Writing patterns	25

Quick facts

Version	v3.0.0
Maintainer	blader (open source)
License	MIT
Install	npx skills add blader/humanizer --global
Works with	Claude Code, Claude Desktop, Skills CLI
What's new	35 patterns merged into 25, in five ranked sections

What is it?

Humanizer v3.0.0 reworks the skill's rulebook: 35 separate AI-writing patterns become 25, grouped into five sections ordered by how strongly each one signals that a machine wrote the text. The project says no checks were lost in the merge — duplicate guidance was folded together, including the rule about dashes, which is now stated once. Humanizer itself is a Markdown skill you install into Claude Code, Claude Desktop or the Skills CLI and call with /humanizer.

How does it work?

The five groups in Humanizer's list are staging instead of stating, rhythm by rule, inflation and borrowed authority, formatting by rule, and leftovers such as chatbot wrappers and knowledge disclaimers. The skill marks what it finds, drafts a rewrite without treating the original structure as fixed, then checks the result against the same list. Only the first five patterns justify an edit on their own; the rest have to appear more than once. A validator in this release counts the patterns from the headings and fails if the skill file grows past 400 lines.

Why does it matter?

Asking a model to write less like a model rarely holds, because the system being asked to hide the habit is the one producing it. A named, versioned checklist applied after the draft is easier to audit and to disagree with. Version 3.0.0 also removes two rules the project decided were wrong — false numeric ranges and synonym cycling are things human writers do too — so fewer honest sentences get flagged, and the watch lists for vague wording, sales language and Markdown formatting got longer.

Who is it for?

writers and developers publishing agent-drafted text

Frequently asked questions

How much does Humanizer cost?: Humanizer is free and MIT-licensed, published as a plain Markdown skill on GitHub. There is no service, account or paid tier attached to it. The cost you do pay is whatever your agent charges for the rewrite itself, since Humanizer runs inside Claude Code, Claude Desktop or another harness that installs skills through the Skills CLI.
Does Humanizer help text get past AI detectors?: Humanizer v3.0.0 explicitly removed the "ai-detection" keyword from its package files, and the project frames the skill as a writing-quality tool rather than a detector-evasion tool. The stated goal is text that reads like a person wrote it, and the README is clear that Humanizer does not invent content while rewriting.
Will Humanizer change the facts in my draft?: No. Humanizer rewrites for voice and structure while preserving the claims in the original text, and the README states plainly that the skill "does not make things up". Patterns it removes are stylistic — staged rhetoric, forced triads, decorative bold, chatbot wrappers — rather than the substance of what you wrote.
What changed in Humanizer v3.0.0 compared with version 2?: Humanizer v3.0.0 is a restructure rather than a feature release: 35 patterns became 25 across five ranked sections, the five-part workflow collapsed into one, and duplicated guidance was merged. The release folds in contributor fixes from eight pull requests, closes three issues, drops two rules that turned out to describe human habits, and reorders the README to put installation first.
Can Humanizer match my own writing voice?: Yes. The README describes giving Humanizer two or three paragraphs of your own writing before the text you want rewritten, so the output follows your style rather than a generic plain register. Without samples, Humanizer still rewrites the draft, but it targets the 25 listed patterns instead of imitating a specific author.

Try it

npx skills add blader/humanizer --global