Agent skill that removes signs of AI-generated writing from text
Install
$
npx skills@latest add drunkrhin0/antislopClaude Code
claude mcp add antislop -- npx skills@latest add drunkrhin0/antislopCursor / Claude Desktop (mcp.json)
{
"mcpServers": {
"antislop": {
"command": "npx",
"args": [
"skills@latest",
"add",
"drunkrhin0/antislop"
]
}
}
}Install command taken from the project's README or registry entry. Always review code before running it.
README excerpt
Read full README on GitHub →Antislop
Antislop helps AI agents write less formulaic prose without losing the author's facts, citations, terminology, or voice.
It provides two skills:
antislopwrites, edits, reviews, and audits prose. It activates when you ask an agent to work on a prose artifact.generate-slopcreates deliberately formulaic samples for demos and tests. It must be invoked by name.
Install
Ask your agent:
Install the antislop and generate-slop skills from https://github.com/drunkrhin0/antislop. Do not install the retired standalone antislop-audit skill.
Or use the Skills CLI:
npx skills@latest add drunkrhin0/antislop
You can also download a package from the latest GitHub release:
antislop-<version>.zipfor the standalone skillantislop-plugin-<version>.zipfor ChatGPT and Codexantislop-kiro-<version>.zipfor Kiro
The plugin package also contains a Claude Code manifest, but Claude Code installation has not been verified.
For ChatGPT for Work, import https://github.com/drunkrhin0/antislop from Workspace settings > Plugins > Add > Import marketplace and leave Path empty. Codex reads the package's Codex manifest. For Kiro, import powers/antislop as a custom Power.
Use
| Goal | Ask your agent |
|---|---|
| Write or edit | Rewrite this release note. Keep every version number and compatibility warning unchanged. |
| Audit without rewriting | Audit this text for formulaic writing patterns. Do not rewrite it. |
| Slop generation | Use generate-slop to create a formulaic marketing sample for an audit test. |
Audits return exact excerpts, rule IDs, and a Formulaic Writing Risk Score:
85-100: low risk65-84: moderate risk40-64: high risk0-39: very high risk
The score measures patterns in the text. It cannot prove or disprove AI authorship. Use generate-slop only for synthetic samples.
Version 3
Version 3 combines writing and audit mode in antislop, expands the registry from 146 to 222 rules, adds five writing profiles, and incorporates 18 reviewed sources. Rewrites now protect facts, citations, links, code, required terms, and the author's voice.
Remove old standalone copies of antislop-audit before installing version 3. Leaving both installed can cause ambiguous routing.
See the release history for detailed changes and sources and credits for attribution.
License
MIT
Similar plugs
Academic Research Skills for Claude Code: research → write → review → revise → finalize
Google Workspace CLI — one command-line tool for Drive, Gmail, Calendar, Sheets, Docs, Chat, Admin, and more. Dynamically built from Google Discovery Service. Includes AI agent…
ARIS (Auto-Research-In-Sleep) — Lightweight Markdown-only skills for autonomous ML research: cross-model review loops, idea discovery, and experiment automation. No framework,…
Self-organizing AI second brain for Obsidian + Claude Code. Drop any source and Claude reads, links, and files it into one connected knowledge graph of plain Markdown you own.…
Free, open-source AI Office suite: Docs, Sheets, Slides, PDF, Markdown and HTML editors with a built-in AI agent, plus a `genoffice` CLI and agent skill so Claude Code, Codex…