Avoid AI Writing

•
4.8k
•
申請済み
管理

Agent skill that audits and rewrites prose to remove AI writing patterns. Rewrite, detect-only and edit-in-place modes, voice profiles and a bundled detector.

avoid-ai-writing is a portable writing skill for Claude Code, OpenClaw, Hermes, Codex, Cursor and other agentskills.io-compatible agents. It audits and rewrites content to remove AI writing patterns, working from a catalog of 74 pattern categories and a 112-entry word replacement table across three tiers. Rewrite mode applies up to two editing passes and returns one final version. Detect mode flags patterns without rewriting and separates clear problems from judgment calls. Edit mode changes a prose file in place with minimal edits and refuses source code, configuration and generated data. An optional voice profile sets how the prose should sound. The repo also ships a zero-dependency detector (npm: avoid-ai-writing-detector) that scores text from 0 to 100, a GitHub Action and pre-commit gate that fail on finding count per file, and a separate MCP wrapper for the detector. The rules and detector are calibrated for English. License: MIT.

主な機能

0174 pattern categories with before-and-after examples
02112-entry word replacement table across three tiers
03Rewrite, detect-only and edit-in-place modes
04Optional voice profiles: casual, professional, technical, warm or blunt
05Zero-dependency detector on npm that scores text from 0 to 100
06GitHub Action and pre-commit gate that fail on findings per file

ユースケース

01Polishing technical blogs, release notes, and documentation to sound human and direct
02Auditing marketing copy, LinkedIn posts, and investor updates to remove synthetic AI phrasing
03Cleaning up AI-assisted drafts directly in markdown or text files without altering code or tables