Who is humanizer for?

Writers, developers and students who use AI to draft text and want it to read naturally; you only need to know how to use an agent such as Claude Code or Codex.

What can you build with humanizer?

Clean up a launch post or blog draft that an LLM wrote before you publish it. Rewrite the prose in a Markdown file such as docs/launch-post.md without touching code blocks or links. Make AI-drafted emails or updates sound like your own writing by supplying a sample. Review your own text against a checklist of known AI tells. Study the 26 patterns to learn what makes writing feel machine-made.

How do I get started with humanizer?

Install the skill for your agent (in Claude Code, add the marketplace and then install the plugin). Then run /humanizer and paste your text, or ask the agent to humanize a file. Commands from the README: /plugin marketplace add blader/humanizer; /plugin install humanizer@humanizer; npx skills add blader/humanizer --global --agent codex.

blader /

humanizer

An agent skill that rewrites AI-sounding text so it reads like a person wrote it

59.7k stars
6.2k stars since 3 Oct
Python
Updated 6 days ago
Beginner friendly
Agent skills & plugins

Open on GitHub (opens in a new tab) Website (opens in a new tab)

AI-written text tends to share habits: "it's not just X, it's Y" lines, one-sentence dramatic closers, forced triads, decorative bold and headings, and leftover chatbot phrases like "I hope this helps!". Readers notice these quickly. Humanizer is a skill you install in an AI agent that spots these habits and rewrites around them.

Its pattern list comes from Wikipedia's "Signs of AI writing" guide. It marks every tell it finds, strongest first, drafts a rewrite, critiques that draft, then gives a final version. It is told not to invent facts: names, numbers, dates and quotes must come from your source, and if a detail is missing it asks you.

You can paste text, point it at a file, or add a sample of your own writing so the result matches your voice. The README says it edits for human readers and is not meant to beat AI detectors, which still flag most of its output.

What it does

26 patterns from Wikipedia's guide
Text is checked against 26 numbered patterns in six groups, such as staging instead of stating, rhythm by rule, inflated significance, formatting by rule and chatbot leftovers. The first five justify an edit on a single sighting.
Draft, critique, final
When you paste text, it shows a first rewrite, a short critique of anything that still sounds artificial, and the final version.
File editing that leaves code alone
Give it a file path and it changes only the prose. Code, data, frontmatter and link targets stay as they are.
Voice matching
Include two or three paragraphs of your own writing and it follows that sample's rhythm, word choice, punctuation and quirks.
Works across agents
It installs in Claude Code as a plugin, in Codex and many other agents through the Skills CLI, and in Claude.ai or Desktop by uploading a ZIP.

What you could build

Clean up a launch post or blog draft that an LLM wrote before you publish it.
Rewrite the prose in a Markdown file such as docs/launch-post.md without touching code blocks or links.
Make AI-drafted emails or updates sound like your own writing by supplying a sample.
Review your own text against a checklist of known AI tells.
Study the 26 patterns to learn what makes writing feel machine-made.

Get started

Install the skill for your agent (in Claude Code, add the marketplace and then install the plugin). Then run /humanizer and paste your text, or ask the agent to humanize a file.

/plugin marketplace add blader/humanizer

/plugin install humanizer@humanizer

npx skills add blader/humanizer --global --agent codex

Commands are copied from the project's README. Read it on GitHub for the full setup.