Humanizer

Humanizer is an open-source writing Skill by Siqi Chen (GitHub user blader), built on Signs of AI writing, the guide Wikipedia editors use to spot AI-generated text. It checks for 26 common AI patterns, such as "not X but Y" contrasts, one-line closers, sayings that only sound deep and staged run-ups like "Honestly?", and rewrites them while keeping what the text says. Once installed in Claude Code, Codex or another agent, call it with /humanizer and paste your text, or point it at a file to edit. The README is clear that it edits for human readers and getting past AI detectors is not the goal; detectors still flag most of its output. The author also reports a blind test in which judges preferred the rewrite 16 times out of 16, which is the author's own test. What sets it apart is that the rules come straight from the hands-on experience of Wikipedia editors, each with a before-and-after example.

Who it's for:writers and editors working on English blog posts, marketing copy, personal or technical writing who want to keep the meaning intact.

Before you start: in Claude Code, install it from the plugin marketplace (version 2.1.142 or later). For Codex and other agents, use npx skills add, a third-party installer that downloads files and writes to your Skill folder, or simply copy SKILL.md. On Claude's web and desktop apps you can download the repository as a ZIP and upload it as a Skill in Settings. There's no dedicated API key, extra fee or auto-publishing; your text goes to whichever AI model you use, under that model's own costs and data policy. It's aimed at English. MIT license.

Based on the project's public documentation. Not tested by AI Inspiration World.