Archestra used Git’s --author flag to fake contributor status and whitelist real humans, after AI bots flooded their bounty issues with 253 comments and 27 untested PRs.
Key Takeaways
GitHub’s “Limit to prior contributors” setting gates issue/PR access on having a commit on main; git commit --author with a user’s noreply email exploits this to grant whitelist status.
Every GitHub account has a deterministic noreply email: <id>+<username>@users.noreply.github.com; the ID is fetchable via gh api users/<username> --jq '.id'.
Archestra built a full onboarding flow: CAPTCHA on their site, a GitHub Action that looks up the user’s ID, commits to EXTERNAL_CONTRIBUTORS.md authored under their account, and pushes to main.
Before this fix, one team member spent half a day per week manually closing hallucinated issues and untested PRs; a prior reputation bot (“London-Cat”) and an “AI sheriff” both failed to stop the flood.
The tradeoff is real: blocking all non-contributors is a “nuclear option” for a VC-backed startup measured on GitHub activity metrics.
Hacker News Comment Review
Commenters flagged a security risk the post overlooks: granting contributor status via the --author trick also bypasses CI approval requirements for fork PRs, since GitHub treats any merged-commit author as a trusted contributor.
The ELO-based reputation idea was quickly countered: any manipulable scoring system will be gamed, and a single AI account that sneaks through can then elevate others, breaking the whole scheme.
Debate split between systemic fixes (GitHub adding per-PR token grants, rejection-rate throttling) and root-cause arguments that bounties and AI hype together created the incentive structure driving the spam.
Notable Comments
@captn3m0: granting contributor status also removes CI approval gates for fork PRs – a privilege escalation path the post does not address.
@krupan: frames the spam wave as a downstream consequence of the industry broadly overclaiming AI coding ability, not just a tooling gap.
