avoid-ai-writing

Why we included this project

Anyone who drafts long-form text with an LLM has felt that moment when a human editor flags a paragraph as 'too AI.' This skill gives a coding agent a concrete way to act on that: a zero-dependency scoring engine flags specific pattern types rather than guessing, and the text is either rewritten or reported back to you. The default mode rewrites the whole piece and runs a second pass to catch patterns the first edit missed. Detect-only mode just lists the issues, which is handy for text you do not want altered, and edit mode makes minimal in-place changes, leaving passages that already read human alone. The optional voice profiles (casual through blunt) keep the tone consistent across output, and since this is a skill rather than a standalone application, it slots into Claude Code and other agentskills.io-compatible agents.