What is Zero Slop?
Zero Slop is a free, open-source agent skill that finds and removes familiar AI writing patterns without losing the draft's core details. It can run in an AI assistant, locally from the command line, or through the browser editor and hosted MCP connector.
Is Zero Slop free?
Yes. The skill and local scorer are free under the MIT license. Local scoring runs offline. Hosted editing is free within shared capacity limits; your AI assistant may apply its own model or account limits.
Does Zero Slop detect who wrote a text?
No. Zero Slop measures surface writing patterns and checks whether an edit preserved source details. Its score is not an authorship probability and cannot establish whether a person or model wrote a draft.
What does Zero Slop check?
It checks stock phrasing, repetition, pacing, readability, formatting, and other documented writing patterns. It also compares names, numbers, quotations, links, code, tables, paths, and structure with the original where the workflow supports those checks.
Does local scoring upload my draft?
No. The local scorer uses Python's standard library and does not send the draft to a server. Hosted rewriting is a separate choice: it sends the draft to the hosted service for editing under the published privacy and capacity terms.
How do I install Zero Slop?
Install the agent skill with `npx skills add manavmishra/ZeroSlop --global`, then use it from an assistant that supports agent skills. If your setup does not support agent skills, use the browser editor at /try/ or connect to the hosted MCP endpoint.
What happens when hosted capacity is exhausted?
The service identifies whether the client allowance or a provider's free capacity is exhausted. When Workers AI capacity is the problem, the service may try another pinned free OpenRouter model. It does not bypass a client's allowance or silently use a paid model. Respect the Retry-After response.
Where can I use Zero Slop?
Use the skill with Codex, Claude Code, Cursor, Gemini CLI, OpenCode, Warp, Zed, and other agents that read SKILL.md files. You can also use the browser editor, MCP connector, REST API, or CLI.