ci: Redline caller pinned to askalf/askalf@35a0ee3 (#62 + #63) - #87
Conversation
sprayberry-redline
left a comment
There was a problem hiding this comment.
Automated review from the Sprayberry Labs fleet code reviewer.
Reviewed by the gating lane (gating review).
Verdict: approve. Pin bump only: .github/workflows/redline.yml:23 moves the reusable workflow reference from 63e007e... to the full SHA 35a0ee3805b3d0ae6df3a7ca63bfebf35a76c6af and updates the trailing comment to askalf/askalf#63. Verified the rest of the file is unchanged: read-only permissions (contents/pull-requests read), pull_request trigger, same-repo and non-draft guard on line 21, per-PR concurrency group, and the runner-label: redline input. The reference is a full commit SHA, so the callee remains immutable. PR text is plain and factual with no AI attribution or secrets. The callee workflow at 35a0ee3 lives in askalf/askalf and is not in this checkout, so its described behaviour changes were not independently read; the caller's security posture does not depend on them.
What
.github/workflows/redline.ymlcallsaskalf/askalf/.github/workflows/redline-review.ymlat a fixed commit. This moves the pin from63e007eto35a0ee3, which carries askalf/askalf#62 and #63.Why
At 18:24Z on 2026-09-25 eight repos ran their first Redline-in-CI job at once and five died with "no review submitted within 30 turns", while dario sat at its concurrency ceiling. The script at
63e007enudged the model 24 times and paid 25 model calls per run for nothing, and logged nothing about what the replies held. At35a0ee3three replies without a tool call end the run, each such reply is logged with its text, an empty reply stays out of the history, and a reply with no content fails at once.Nothing else changes
Same runner label, same guard, same permissions. The workflow is a pin bump only.