Post

🏔️ Peak
@peaklabs.dev

Introducing Anti Slop 🚫 A GitHub action that detects and automatically closes low-quality and AI slop PRs. As one of the maintainers of Coolify (50K+ stars) we deal with slop PRs daily so I decided to fix it. Early testing showed this action could have closed 98% of slop PRs.

🏔️ Peak
@peaklabs.dev

◆ The entire action runs in under 15 seconds which means slop PRs are caught and closed before you even notice them ◆ 22 check rules (more coming) with thoughtful defaults derived from patterns identified across 100+ manually reviewed AI slop PRs ◆ Configurable sensitivity to avoid false positives

🏔️ Peak
@peaklabs.dev

◆ Owners, Members and Collaborators are automatically exempt by default with zero configuration needed ◆ The beauty of open source is that anyone can contribute. Anti Slop blocks low-quality and AI-generated PRs without forcing you to restrict contributions to Collaborators only

seancj.bsky.social
@seancj.bsky.social

Sounds like a semi-decent solution for now, but also sounds like a stopgap measure. Agents are going to be pointed at these kind of things, and then make the appropriate adaptations to their PRs for them to not trigger the filter.

🏔️ Peak
@peaklabs.dev

Maybe in the very long run but for now it works very well for @coolify.io and some other big projects are also using it. Some checks will also always work, such as the user's merge ratio, spam patterns, amount of forks in 24h, account age, etc. There will be even more checks in v0.4 coming soon.

0 reposts3 likes

seancj.bsky.social
@seancj.bsky.social

Hey, I'm all for looking for solutions, and as I said, I don't doubt this works (for now)! I guess my larger point is this is just one more negative the LLM wave is bringing to our world. (And yeah, kinda worried that even the points you make are gameable - though at least they *will* slow agents)