Humanizer
ProductivityRemoves signs of AI-generated writing from text, based on Wikipedia's detection guide.
Install
$ git clone https://github.com/blader/humanizer ~/.claude/skills/humanizer
Update
$ git -C ~/.claude/skills/humanizer pull
Or re-clone/re-copy from the repo if it is not a git checkout.
When to use it
- Editing README files, blog posts, or docs drafted with AI assistance
- Marketing copy that needs to sound human, not generated
- Any text where AI tells (inflated symbolism, rule of three, em-dash overuse) would hurt credibility
Commands & flags
- Humanize the README I just wrote
- Primary invocation: ask Claude to humanize a finished text.
- Use humanizer on docs/announcement.md
- Run the skill against a specific file before publishing.
- Review this draft for AI tells
- Detect promotional language, vague attributions, and AI vocabulary.
- Fix the em-dash overuse and rule-of-three
- Target specific patterns like negative parallelisms and filler phrases.
How to use it
Ask Claude to humanize a text or review a draft:
Humanize the README I just wrote
The skill detects and fixes specific patterns from Wikipedia's "Signs of AI writing" guide: promotional language, superficial -ing analyses, vague attributions, negative parallelisms ("not just X, but Y"), AI vocabulary words, filler phrases, and em-dash overuse.
Why it earns its place
Readers have learned to spot generated text, and it costs trust. This skill encodes a concrete, sourced checklist instead of a vague "make it sound natural" instruction, so the edits are systematic and explainable. Especially useful before publishing anything public-facing.
Try it
> Use humanizer on docs/announcement.md before I publish it