avoid-ai-writing is a skill that audits and rewrites content to remove AI writing patterns, designed to work with AI agents including Claude Code, OpenClaw, Hermes, and other agentskills.io-compatible platforms.
The tool addresses the problem of AI-generated text containing recognizable patterns—verbose phrasing, inflated language, generic conclusions, and other markers of machine writing. It works by scanning content for these patterns and either flagging them for review or rewriting the text to sound more human. The approach includes a built-in second pass to catch patterns that survive the initial rewrite, and supports three distinct modes: Rewrite flags problems and fixes them automatically, Detect identifies patterns without altering text, and Edit modifies prose files in place while refusing to touch source code or structured data that could be corrupted.
Developers should choose this tool if they need to audit AI-generated content before publication or want an agent to clean up its own output. It suits projects where maintaining human voice matters—marketing copy, documentation, or editorial content—but not where AI patterns are intentional or where the text is already structured data. The tool offers voice profiles to control tone independently of audience context, and an iteration option to repeat the audit cycle up to a second pass if needed.
The project shows active development with recent commits addressing core functionality. The codebase demonstrates attention to practical edge cases, such as refusing to rewrite structured content and providing detailed reports on what was changed and why. The tool includes a concrete example showing how it transforms verbose startup marketing language into clear, factual statements while preserving specific details like funding amounts and product capabilities.
