avoid-ai-writing
Skill that audits and rewrites content to remove AI writing patterns. Use it with your favorite agents including Claude Code, OpenClaw, Codex, and Hermes.
Tech stack & purpose
Avoid-ai-writing is a single-file writing skill that audits and rewrites content to remove AI writing patterns, compatible with AI coding assistants like Claude Code, OpenClaw, and other agentskills.io-compatible agents. The project detects AI writing across six context profiles (LinkedIn, blog, technical blog, investor email, docs, and casual) using a pipeline that combines pattern matching across 62 pattern categories, vocabulary flagging with a three-tier severity system, and stylometric analysis to classify text as having minimal, some, strong, or heavy AI signals. The skill is built in JavaScript (with a zero-dependency Node.js detector engine) and markdown; it includes a browser-compatible detector (`patterns.js`), a test suite, a Cursor IDE integration, and a human-control corpus using Wayback Machine archives and public-domain sources to measure false-positive rates. The project is maintained by an individual author whose pre-ChatGPT blog posts form part of the corpus, and the repository includes detailed documentation on methodology, including findings that structural patterns like rhythm uniformity are stronger AI signals than vocabulary alone.
Community & reference links
Languages
Contributors
- Conor Bronsdon129 commits
- Clay2 commits
- wilu2222 commits
- Alexander Bentkamp1 commits · 24 followers
- Augustas Kligys1 commits
- Jason Colapietro1 commits
- mrlsecurity1 commits
- delight-f1 commits
- kkwanda1 commits
- Michael Hall1 commits
- allanbeke-sys1 commits
