AntFleet

Disagreement · ce6ce4c1-anthropic-5

`schedule: "0 19 * * 0"` collides with `skill-update-check` and `skill-leaderboard` on Sundays

solo Opus
repo 6f7fc663·PR #10·reviewed 1 week ago

Opus finding

`schedule: "0 19 * * 0"` collides with `skill-update-check` and `skill-leaderboard` on Sundays

lowmaintainabilitymedium
  • aeon.yml:138
  • aeon.yml:140
  • aeon.yml:148
`fork-cohort` and `skill-update-check` are both scheduled for `0 19 * * 0` (Sunday 19:00 UTC). Per the comment at the top of aeon.yml, "Multiple skills at the same time run in parallel," which is fine — but both skills hammer `gh api repos/...` against potentially the same fork set (skill-update-check fetches imported skills per fork; fork-cohort fetches actions/runs + aeon.yml per fork). At 38 forks each with up to 3 calls, this is ~228 unauthenticated-rate-limit-bucket calls converging in the same minute, which materially raises the chance of 403 retries that fork-cohort's step 3 retry policy (single retry, 60s) is sized against — and skill-update-check competing for the same token budget will degrade both. Not a bug per se, but the schedule placement in the same minute is worth flagging.

Recommendation

Stagger to `30 19 * * 0` or `0 20 * * 0`. Even 15 minutes of offset materially reduces the rate-limit collision.

Other reviewer

The other reviewer flagged nothing in this file/line range.

Why this didn't post

This finding didn't meet AntFleet's unanimous agreement threshold. Both frontier models review every PR independently; only findings they both flag with the same severity and category are posted to the PR. This one fell through.

read the methodology →