Humanizer
blader/humanizer
A single skill that rewrites AI-sounding text so it reads like a person wrote it, without changing what it says.
Humanizer is a free, MIT-licensed agent skill by Siqi Chen that removes signs of AI-generated writing from text. It checks for 26 patterns taken from Wikipedia's Signs of AI writing guide, rewrites the passage and keeps the facts as they were. It works in Claude Code, Codex and other agents that support skills, and the README says getting past AI detectors is not a goal.
Facts checked on against the sources listed at the end of this page.
- GitHub stars
- 53,568
- Skills in the pack
- 1
- Agents supported
- 6plus any agent the skills CLI supports
- Instruction size
- 5.2k wordsacross its SKILL.md file
How to install Humanizer
| Agent | Command | Copy |
|---|---|---|
| Claude Code | Needs Claude Code 2.1.142 or newer. The plugin answers to /humanizer:humanizer. | |
| Codex | | |
| Gemini CLI, GitHub Copilot, Windsurf and other agents | Installs for every agent the skills CLI supports. Leave off --global to install only in the current project. | |
| Claude.ai and Claude Desktop | Download the repository as a ZIP and upload it as a skill in Settings. |
Commands as given in the README on 2 October 2026.
What is inside Humanizer
| Skill or command | What it makes the agent do |
|---|---|
| humanizer | Marks every sign of AI writing it finds, strongest first, drafts a rewrite, checks the draft against the patterns and the original claims, then writes the final version. |
One skill that checks 26 patterns in six groups: staging instead of stating, rhythm by rule, inflation and borrowed authority, formatting by rule, leftovers from the chat, and writing for the wrong reader.
What Humanizer runs on your machine
| Hooks | No. |
|---|---|
| Bundled scripts | No. The skill is a single SKILL.md of instructions. The repo also holds a validate-package.py script that the skill does not run. |
| Network calls | No. |
| API keys needed | None. |
These are facts read from the repo, not a security rating. Read the files before you install any skill.
How much context Humanizer uses
| The skill | 5,219 words in its SKILL.md file |
|---|---|
| Loaded at start | Only each skill's name and description. The full text loads when a task needs that skill. |
Word counts are measured from the repo every day. Reference files a skill loads on demand are not counted.
How Humanizer works
Humanizer is built on the guide Wikipedia editors use to catch AI-generated text. The README names five patterns it rewrites on a single sighting: not X but Y contrasts, one-line closers, sayings that sound deep, a staged run-up before the point, and arguing with no one. Other patterns, such as dashes used as a universal connector or curly quotation marks, count only when several appear in the same passage.
You call it with /humanizer and paste text, ask in plain language, or give it a file path. For pasted text it shows a first rewrite, a short critique and the final version. Given a file, it changes only the prose and leaves code, data, frontmatter and link targets alone. It does not invent details: if a sentence needs a fact that is missing, it asks. You can paste a sample of your own writing and it follows that sample's rhythm and word choice.
A good fit for
- Editing drafts an AI model wrote before a person reads them
- Cleaning the prose in docs and READMEs while leaving code blocks untouched
- Matching your own voice by giving it a sample of your writing
- People who want a skill with no hooks, scripts or network calls
Skip it if
- Getting past AI detectors: the README says that is not a goal and that detectors still flag most of its output
- Adding substance to thin text, since it keeps the claims of the original and does not make things up
- Claude Code versions older than 2.1.142, which need the skills CLI install in place of the plugin
Is Humanizer still maintained?
| Last commit | , 4 days ago |
|---|---|
| Latest release | v3.1.0, 28 September 2026 |
| Stars gained in 7 days | Shown after 7 days of tracking |
| Stars gained in 30 days | Shown after 30 days of tracking |
| Open issues and pull requests | 4 |
| Forks | 4,268 |
Tools Humanizer works with
Questions about Humanizer
Does Humanizer bypass AI detectors?
No. The README says Humanizer edits for human readers, that getting past AI detectors is not a goal, and that detectors still flag most of its output.
How do I install Humanizer in Claude Code?
Run /plugin marketplace add blader/humanizer, then /plugin install humanizer@humanizer. It needs Claude Code 2.1.142 or newer. On older versions use npx skills add blader/humanizer --global --agent claude-code.
Does Humanizer work with Codex and Gemini CLI?
Yes. For Codex run npx skills add blader/humanizer --global --agent codex. The same command with --agent '*' installs it for every agent the skills CLI supports, including Gemini CLI, GitHub Copilot and Windsurf.
What patterns does Humanizer look for?
It checks 26 patterns based on Wikipedia's Signs of AI writing guide, including not X but Y contrasts, one-line closers, forced lists of three, overused AI words, inflated significance, bold used as decoration and leftover chatbot phrases.
Is Humanizer free?
Yes. It is open source under the MIT license.