Thanks for helping make AI writing less detectable. Here's how to contribute.
- Open
banned-vocabulary.md - Add the word/phrase to the appropriate table
- Include a plain alternative in the "Use instead" column
- In your PR description, link to evidence — either the Wikipedia article or published research showing the word is overrepresented in AI text
Don't add words that are perfectly normal in human writing. The bar is: would a Wikipedia editor flag this as an AI tell?
- Add a "bad" example to
examples/bad-examples.mdwith annotations for every AI pattern present - Add a matching "good" rewrite to
examples/good-examples.mdwith a "Why this works" section - Examples should come from realistic contexts: project descriptions, emails, docs, blog posts, PR descriptions
New rules should be backed by the Wikipedia article or equivalent evidence. Open an issue first to discuss before submitting a PR — this keeps the rule set focused.
If you use an AI tool that isn't covered in agents/, add a new file with install and usage instructions. Follow the format of the existing files.
- One rule, one example set, or one agent per PR
- Keep changes focused
- Test that your additions don't break existing SKILL.md formatting
- No promotional language in your PR description (practice what we preach)
Be direct, be helpful, don't waste people's time. That's it.