AI Slop PR Guard
Version updated for https://github.com/Zensoro/ai-slop-detector to version v1.0.0.
- This action is used across all versions by ? repositories.
Action Type
This is a Composite action.
Go to the GitHub Marketplace to find the latest changes.
Action Summary
This GitHub Action uses pure regular expression structural features to score PR/Issue bodies, flagging suspected AI-generated content with a label and scored comment. It labels without automatically closing issues unless opted in. The action is designed to reduce false positives by subtracting human signals such as referencing issue numbers or short bodies.
What’s Changed
v1.0.0 — AI Slop PR Guard
The $0 AI-slop guard for open-source maintainers. Pure regex, zero dependencies, label-only by default.
What it does
- Scores PR/Issue bodies on open; flags probable AI-generated slop with a label + scored comment
- Never auto-closes unless you opt in; members/bots never flagged
Reliability
- API retry/backoff on 429/5xx, graceful non-fatal error handling
- Short bodies (<20 chars) skipped
Accuracy
- Human signals (issue references) subtract from the score — fewer false positives
Docs
- Real demo screenshot, FAQ, bilingual README
- Issue/PR templates + GitHub Sponsors funding
Signals (measured on GH Archive 2025-07/08)
| Signal | AI PRs | Human PRs |
|---|---|---|
## Test headers | 78.3% | 0.8% (~98x) |
## section headers | 95.1% | 8.2% (~11.6x) |
| ✅ emoji | 17.9% | 0% |
Install
- uses: Zensoro/ai-slop-detector@v1