avoid-ai-writing

What avoid-ai-writing does

avoid-ai-writing works in three modes. Rewrite, the default, flags AI patterns, applies up to two editing passes, verifies the result when possible and returns one final version. Detect flags patterns without changing the text and separates real problems from judgment calls. Edit changes a prose file in place with minimal, targeted edits and refuses source code, configuration and generated data, because prose rewrites could corrupt structured content.

An --iterate 1 or --iterate 2 option sets the editing-pass ceiling, and an optional voice profile (casual, professional, technical, warm or blunt) controls how the prose should sound. The README organises its pattern reference into content, language, structure, communication and meta patterns, with later additions for structural detection and share-post framing. A bundled detector is published on npm as avoid-ai-writing-detector and can score a file or piped text from the command line. The rules and detector are calibrated for English.

Key features

• Rewrite, detect and edit-in-place modes
• Up to two editing passes, set with --iterate 1 or 2
• Voice profiles: casual, professional, technical, warm and blunt
• Edit mode refuses source code, configuration and generated data
• Bundled detector published on npm as avoid-ai-writing-detector
• Pattern reference covering content, language, structure and communication

When to use it

• Clean up a draft blog post or email that reads like AI output
• Scan a document in detect mode without altering it
• Edit a Markdown file in place while keeping passages that already read as human

Who it is for: Writers and developers who draft prose with AI agents and want to remove typical AI-sounding patterns.

How it fits with Hermes Agent

Supports Hermes explicitly; the README lists Hermes alongside Claude Code and OpenClaw and includes a Hermes section in its installation guide.

FAQ