Skip to content

Humanizer

Rewrites AI-generated text to read more like a person wrote it, keeping the meaning.

Stars
63.8k
Forks
5k
Licence
MIT
Language
Python
Last update
9 days ago
Type
AI model / agent

What it is

Humanizer is an agent skill that takes text written by an AI and rewrites it to remove common signs of machine writing, such as stock phrases, dramatic one-line closers, and rigid formatting. It is based on Wikipedia's guide to signs of AI writing and works in Claude Code, Codex, and other agents that support skills. The README says it changes style without changing what the text says, and that it is not designed to fool AI detectors.

  • Rewrite a pasted draft of a launch post to remove hype and filler
  • Run it on a file path such as a Markdown document, changing only the prose
  • Rewrite text to follow the rhythm and word choice of a sample of your own writing

Great for

  • Writers who draft with an AI and want the final text to sound more natural
  • Developers using Claude Code or Codex who want a reusable writing-cleanup step
  • Users who want rewrites that match a sample of their own voice
  • Editors reviewing docs or posts for stiff, inflated AI phrasing

Not ideal for

  • People who need reliable AI-detector evasion, since the README says detectors still flag most of its output
  • Anyone who expects it to add or verify facts, since it is meant to ask for missing details rather than invent them
  • Users of agents that do not support skills, unless they manually copy SKILL.md into the agent
Step-by-step guide: how to use HumanizerInstall, first run, examples, tips and FAQs

How to get it

  1. 1

    Install with a package manager

    You'll need: Claude Code 2.1.142 or newer for the plugin install path (older versions use npx skills) · Node.js and npx for the Skills CLI commands

    Install for Codex

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

    Install for all supported agents

    npx skills add blader/humanizer --global --agent '*'
  2. 2

    Run it

    • Install the skill using the command for your agent.
    • Type /humanizer followed by the text you want to rewrite.
    • Or ask in plain language, such as 'Please humanize this text: ...'.
    • Review the first rewrite, the critique, and the final version the skill shows.
  3. 3

    Get the most out of it

    • Include 2-3 paragraphs of your own writing after /humanizer to match your voice.
    • Point it at a file path to rewrite only the prose while leaving code and frontmatter alone.
    • Read the 26-pattern reference in the README to understand what it changes and why.

Downloads

Also on GitHub as mirrors or translations (not listed separately): op7418/Humanizer-zh

Strengths

  • Built on a published reference of AI writing patterns from Wikipedia editors
  • Shows its first rewrite, a critique, and the final version when given pasted text
  • Installs through the skills CLI for several agents from one command

Trade-offs

  • The README says AI detectors still flag most of its output
  • Its effectiveness claims rest on a single linked blind test, which the README reports as 16 of 16 preferences
  • Installation and use depend on the particular agent's skill system, and the README is truncated in places

How it compares

ProjectStarsTypeLevelUpdatedBest for
humanizer (this)63.8kAI model / agentbeginner9 days agoWriters who draft with an AI and want the final text to sound more natural
9router31.8kDev toolintermediate9 days agoDevelopers who use AI coding tools and want to reduce spending
planning-with-files27.8kDev toolbeginner6 days agoDevelopers running long multi-step tasks in Claude Code or Codex CLI
hermes-agent260kAI model / agentintermediate4 days agoDevelopers who want a personal agent they can host on their own server or machine
ponytail182kAI model / agentbeginner8 days agoDevelopers who use Claude Code or similar AI agents and find the generated code too long
agent-skills113kOtherbeginner4 days agoDevelopers who already use an AI coding agent and want more structured, test-driven output

Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.