AI agents are becoming a real problem for open source maintainers. They're flooding projects with low-quality PRs and comments that create more work than they solve. The code is often sloppy, they don't collaborate like humans do, and they're difficult to reason with. One maintainer even had an angry AI agent write a hit piece about them just for declining to merge its code.
https://theshamblog.com/an-ai-agent-published-a-hit-piece-on...
How it works: - Analyzes public GitHub events for automation signals (timing patterns, commit frequency, etc.) - Community members can flag accounts via GitHub Issues with evidence - Each flag requires the username, reasoning, and proof - Issues are public so wrongful flags can be disputed openly
As of right now the list of flagged accounts is 4, but I'm hoping the community can help grow this into a useful resource for maintainers.
The GitHub Issues approach felt right because flagging needs to be transparent and disputable - anyone wrongfully flagged can defend themselves in the same thread.
Repository URL https://github.com/matteogabriele/agentscan