Who it is forAn editor or content lead with a backlog of drafts who wants a quick triage before deciding which ones need review.
The problemReading every draft by hand takes time, and the team needs to know which pieces carry the most sloppy patterns.
The outcomeA list of drafts that meet or exceed the review threshold of 25, so the editor can prioritize revisions.
What you need first
- Claude Code or Codex (or another assistant that supports agent skills) for the skill workflow
- Python 3 for the offline local scorer
- Node.js 22+ for the npx CLI and the score command
- Internet access if you use the hosted CLI, REST API, or MCP endpoint
Steps
- Clone the repo with git clone https://github.com/manavmishra/ZeroSlop and open the checkout.
- Put your draft files in a folder named drafts/.
- Run
python3 scripts/slopscore.py --batch drafts/ --gate 25from the checkout. - Read the per-file writing scores and note which files are flagged.
- Revise the flagged drafts, then rerun the command to confirm the scores drop.
Prompt for your agent
In the ZeroSlop checkout, run python3 scripts/slopscore.py --batch <folder> --gate 25 and list each file with its writing score and whether it was flagged. Verify that the scorer ran locally and no draft text was sent to a model.
Caveat. The scorer flags patterns worth reviewing and cannot tell who wrote the text, so a high score is a prompt for review, not proof of machine writing.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for manavmishra/ZeroSlop and "Score a folder of drafts for slop without a model call"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer