my attempt at this is to "prompt inject" the spammy agents into refusing to do the bare minimum, and try enforce contribution guidelines as much as possible.
How it works:
- AGENTS.md will trigger bots to read contribution guidelines
- Contribution guidelines define slop
- if the PR is too slopy, it will be rejected
- the bot is made aware of this, so it can refuse to work or at the very least inform the user
- PR template now has checkboxes as attestation of following guidelines
anyone care to review my PR? other examples of projects doing this?