Repository navigation
feat(ci): flake-inputs-update escalates failures to Jules, then Muse tasks - #1749
Conversation
…tasks Drew 2026-10-28: automated update failures should make an issue and call a Jules; if the Jules fails, Muse tasks triage/fix. The workflow already maintained the single tracking issue; this adds the escalation step after it: attempt 1 dispatches a Jules session (AUTO_CREATE_PR) against the tracking issue via the Jules API using the existing JULES_API_KEY secret, and comments the session link (or the no-secret/no-session fallback for Muse tasks); attempt 3 comments the issue for the Muse repo tasks to take over. Header comment documents the chain. YAML validated; the embedded bash passes bash -n.
|
Warning Review limit reachedYou've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. Next included review available in 34 minutes. View limit details
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
Implements Drew's escalation chain for automated update failures (2026-10-28): issue → Jules → Muse tasks.
The workflow already maintained the single tracking issue (open on first failure, edit in place, close on success). This PR adds the missing two legs, as one new step — Escalate the failure (Jules first, then Muse tasks) — running after the failure is recorded:
AUTO_CREATE_PR, starting branchmain) using the repo's existingJULES_API_KEYsecret, prompted to fix the tracking issue and deliver a PR withFixes #<n>. The session link is commented on the issue. If the secret is missing or session creation returns nothing, the comment says so and hands the issue to the Muse repo tasks instead — the run never fails on escalation.The header comment documents the chain.
Verification: YAML parses; step order is record → escalate → publish; the embedded bash passes
bash -n. The failure path can't be exercised end-to-end without a real failure — the next genuine failure will be its first live test, and every branch of the step degrades to an issue comment rather than a broken run.Need help on this PR? Tag
@codesmith-botwith what you need.