Zero Slop: a skill against AI writing patterns
Zero Slop
A free skill that finds AI writing patterns in a draft, guides the edit and checks that the author's facts and wording survived
- 167 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The skill only reads and edits your text; network use is minimal (update check) and can be switched off
- The hosted MCP and the deslop command send a draft to a remote service
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/zero-slopDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skill only reads and edits your text; network use is minimal (update check) and can be switched off
- The hosted MCP and the deslop command send a draft to a remote service
This is third-party code. Review the repository files before installing.
What it does
Zero Slop runs inside your assistant (Claude Code, Codex, Cursor) and starts with `/zero-slop`. A local Python script scores text from 0 to 100 and finds specific patterns: contrast formulas, canned openers, vague attribution to "experts", inflated significance, promotional wording and uniform sentence rhythm. The score uses 294 weighted patterns and a 96-term lexicon. The assistant itself edits through eight stages: scoring, reading intent, rewriting, a fact check, copy editing, a read-aloud pass, comparison with the original and a final pass. The stages are roles, not separate models. The fact check runs locally: names, numbers, quotations, links, code, tables and paths must survive the edit. `/zero-slop inspect` shows the flagged passages without editing. There are three ways in: the skill, a browser editor at zero-slop.ai and a hosted MCP connector at https://mcp.zero-slop.ai/mcp. The author talks about private learning: the skill remembers your edits and reasons, with data kept under `$ZERO_SLOP_HOME`.
Who it is for. Writers and marketers who draft with AI and want to strip template phrasing from the text without losing facts.
Good fit when
- A LinkedIn announcement came out smooth but sounds like a typical chatbot answer
- Before sending a report you want to see which phrases a reader will take as clichés without retyping the text
- You need a check that numbers, names and links did not get lost in the edit
- You want to train the assistant to your style preferences
Not a fit when
- The text is in Russian: the patterns and lexicon are written for English, and the README does not mention other languages
- You need to prove authorship or detect machine text: the score finds patterns but does not identify an author
- You need an edit with no outside service: the npm `zero-slop deslop` command and the hosted MCP send text to a remote server
Example request
/zero-slop inspect: show which phrases in this announcement sound like an AI template and give the score.Limitations
The score finds patterns but does not identify the author, and your assistant does the editing, so quality depends on the model. The scoring script and pattern descriptions are written for English, and the README does not claim Russian support.
The skill may call GitHub's public releases API to check for updates without sending text. Turn this off with `ZS_NO_UPDATE_CHECK=1`. The hosted MCP and REST send a draft to the Zero Slop service under a separate privacy policy. The separate cleanup of invisible characters does not guarantee removal of statistical model watermarks.
How to disable. Delete the skill folder from the agent's skills directory (with the --global flag that is ~/.claude/skills/ or equivalent). Learning data sits in `$ZERO_SLOP_HOME` and can be erased by hand.
FAQ
How do I install Zero Slop in Claude Code or Codex?
Run `npx skills add manavmishra/ZeroSlop --global`, then call `/zero-slop (your text)`. For Claude.ai upload the zero-slop.zip archive from the latest release. Gemini CLI and remote MCP connections have separate instructions in DISTRIBUTION.md.
What is workslop?
The README uses the word for smooth but empty work text produced by AI with no thought behind it. The skill looks for its signs: clichés, inflated significance and evasive references to "studies".
Can I check a text without a model?
Yes, the local score runs as a script with no model call: it counts points and lists the phrases found. Editing and rewriting are done only by the assistant attached to the skill.
Does my text go to a Zero Slop server?
The installed skill runs in your AI assistant with local checks, with no separate model or service. A draft goes to a remote service only when you explicitly call the npm `zero-slop deslop` command or connect to the hosted MCP and REST. The update check against GitHub's public API is turned off with `ZS_NO_UPDATE_CHECK=1`.
Does Zero Slop detect who wrote a text?
No. The score finds patterns worth reviewing but cannot identify an author, and the authors warn that detectors misclassify non-native English writing.
Discussion
No questions or reviews yet.
Related
SkillsA skill that finds AI tells in English prose and edits them surgically; three Python scanners run on their own and the teach command learns your voice
A prose-checking plugin that runs a local detector after supported file saves and gives the agent editing feedback
A skill that edits English drafts: strips AI boilerplate, keeps the author's voice, and can also just list what it found without rewriting
Two anti-slop skills: slopbeth edits writing you publish and slopgent keeps the agent's own replies honest and plain.
A writing rules skill for agents: it strips model boilerplate, varies sentence rhythm and bans invented numbers and quotes
Humanizer (brandonwise)
humanizer
A skill and Node.js CLI that score English text for AI-writing patterns and statistics and suggest fixes on a 0 to 100 scale
Comparison with similar tools
Cards that share the main search query, and how they differ.
README in short
The README shows an edit example: an announcement scores 99.3 out of 100 and 9.5 after processing. It then describes the patterns, eight stages, fact gate, learning from your edits, limits of the score and ways in through the skill, browser and MCP.
README badge
Original SKILL.md
--- name: zero-slop license: MIT metadata: version: "2.12.38" author: manavmishra description: Edit drafts into natural prose, inspect AI-sounding patterns, or review how a specified audience might respond passage by passage. Zero Slop runs inside the user's existing AI assistant with local tools that protect source details. Use for humanizing or de-slopping writing, polishing outward-facing prose, social drafts, final editorial checks, or an explicit simulated reader review. Preserve facts, voice and format; reader simulations are hypotheses, not human feedback. --- # Zero Slop A linter for the AI accent. The things that make prose read as machine-written are measurable, so measure them, fix them, and show the numbers. Zero Slop is a skill, not an AI model. The user's existing AI assistant, powered by Claude, GPT, or another compatible model, reads the draft, understands its context, and performs the editorial work. The bundled local tools handle repeatable checks. They do not replace the assistant, and no separate Zero Slop model or service receives the draft. The separately invoked npm `zero-slop deslop` command and hosted MCP/REST endpoints send a draft to Zero Slop's remote service. They are opt-in alternatives, not local checks in this workflow. Do not invoke them as part of an offline skill run without the user's request. The npm `score` command continues to run locally. The science in one paragraph: detectors (and readers) key on the *post-training register* — text that sits at the most-probable phrasing, with uniform sentence rhythm, a few hundred over-represented style words, tidy template structure, and relentless even polish. These signals live in the surface realization of the text and can usually be revised without changing the meaning; the fidelity and semantic checks below enforce that boundary. `references/evidence.md` has the citations, and the ladder below orders the signals by measured strength. ## Hard rules (non-negotiable) For an explicit character-cleanup request, use the bundled offline text-hygiene helper at `scripts/text_hygiene.py` when available. It inspects formatting characters and removes only a small reviewed set of embedded artifacts in 16 common ASCII prose words. Only sentence-shaped lines with conservative prose openers are eligible. Recognized code, command syntax, URLs, formulas, quotations and multilingual formatting are excluded; these guards are not a complete code parser. Preserve uncertain text and use inspection when its purpose is unclear. This is a separate local operation, not a replacement for contextual editing. Its change record can reverse the exact cleanup. A scriptless assistant must not claim byte-level cleanup it did not perform. Character cleanup does not verify removal of statistical watermarks from ChatGPT, Claude, Gemini or any other model. Preserve required disclosures and attribution.