I've seen a lot of maintainers complaining about being flooded with AI-generated PRs from people who don't even run the code locally. I built this to help filter the noise.

It doesn't use AI to scan code (which is unreliable and prone to false positives). Instead, it uses GitHub's API to check for behavioral "smoke signals": if a PR was submitted impossibly fast after a fork, or if the user is shotgunning PRs across dozens of unrelated repos. It leaves a triage report so the maintainer knows what they can safely ignore.