Avoid AI Writing
Audit prose for machine-sounding patterns and rewrite it while keeping the author's voice.
- Repository
- wshobson/agents
- Stars
- 42,095
- Licence
- MIT
- Files
- 4 files
What it does
When installed, the agent can review a piece of writing such as a README, changelog, release note, PR description, or blog post and point out phrases and structures that read as machine-generated. It can list the flags only (detect mode), return a cleaned-up rewrite with a summary of changes (rewrite mode), or edit a prose file in place with minimal targeted changes (edit mode). It uses reference files that grade the severity of patterns, a tiered vocabulary list, and context profiles such as LinkedIn, docs, or investor email that adjust how strictly the rules apply. The skill is explicit that flags are only writing-quality signals. It warns that AI detectors are unreliable, so flags should not be used to decide academic integrity, hiring, or attribution questions. It also forbids adding invented personality, facts, or numbers during a rewrite.
Great for
- Developers polishing READMEs, changelogs, release notes, or PR descriptions before publishing
- Marketers and bloggers who want drafts to sound less generic
- Non-native English writers checking that their own voice comes through clearly
Skip it if
- Academic-integrity, hiring, or attribution decisions, because the skill's own documentation says flags are not evidence of AI use and detectors have high false-positive rates
- Source code, configuration, or generated data files, which the edit mode explicitly refuses
- Quoted, attributed, or third-party text that the author does not own, where detect mode is safer than rewriting
Try asking
- “Audit this blog post draft for AI-isms and list which flags are clear problems and which are judgment calls”
- “Rewrite this release note so it sounds less machine-generated, using the docs context profile”
- “Clean up the README at ./README.md in place and confirm the flagged patterns are gone”
- “Scan this LinkedIn post for AI tells, then rewrite it in a casual voice”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 4 original files, the MIT licence and an attribution note.
avoid-ai-writing.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip avoid-ai-writing.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip avoid-ai-writing.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Use detect mode first if the text is published or belongs to someone else, so you decide each change yourself
- Choose a context profile explicitly (for example --context docs or --context linkedin) so the rules match the genre
- After a rewrite, check that the voice still sounds like you; the skill warns against adding stance, anecdotes, or facts the source never had
What's inside
- references/pattern-catalog.md
- references/profiles.md
- references/word-tiers.md
- SKILL.md
Redistributed unmodified from wshobson/agents under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.