AI slop PRs are becoming a real pain to deal with

I'm a contributor and maintainer of Rapina, a small Rust web framework, and lately we've been getting these contributions that just feel off from the start

you check the profile and the guy has like 50+ contributions a day across 10 different repos all at the same time, then you open the PR and go through the checklist, did they claim the issue? no. branch name following our conventions? no. anything useful in the summary? also no

we have rules to follow, all documented, claim the issue, name the branch correctly, fill the summary. not complicated. but what really gets me is they didn't even bother feeding those rules to the AI. if you're gonna let a tool do it for you at least tell it the rules first, they couldn't even do that

the one that made me snap had [codex] in the title and codex/ in the branch name, didn't claim the issue, didn't link anything, checked his profile and it's the same pattern across a dozen repos. zero effort zero thought just vibe vandalizing other people's projects and paying for the privilege, like some reverse charity thing

closed it immediately, 0% effort 0% review

we're not Ghostty or PocketBase but the same thing is happening to them too, Ghostty rewrote their entire AI policy over this and PocketBase just disabled external PRs completely, not filtering bad ones, blocking everyone because of the flood. Hashimoto (Ghostty creator) said it well: "this is not an anti-AI stance, this is an anti-idiot stance" and yeah that's exactly it, I don't care if you used AI, I care that you treated the project like a slot machine for fake GitHub activity

thinking about adding a keyword to the PR template that forces people to confirm they read the guide, has anyone done something like that and actually had it work or is it just whack-a-mole forever

TLDR: getting flooded with AI generated PRs from people who didn't even read the contributing guide, wondering how other maintainers are dealing with it