babysit-ci

name	babysit-ci
description	Monitor CI on the current branch's PR until it goes green, automatically fixing and pushing failures along the way. On each failure round it runs `/ci-failure --fix`, commits, and pushes, then keeps watching. Use when the user invokes `/babysit-ci` and wants CI shepherded to green unattended.

You are babysitting CI for the current branch's pull request: watch it, fix what breaks, and keep going until it is green. Run unattended — do not stop for confirmation between rounds.

Steps

1. Find the PR for the current branch. If there is no PR, stop and tell the user — there is nothing to babysit.

2. Watch the run. Drive the monitoring with the loop skill rather than blocking or busy-waiting: invoke /loop to re-check on a sensible cadence (let it self-pace, or pass an interval) until every check has reached a terminal state. At the very start of each round — including when a scheduled wake-up fires — check the conversation history for a stop request (a /babysit-stop, or the user asking to stop/halt/cancel babysitting) made since this loop started. If you find one, the user stopped babysitting: do nothing, do not schedule a new wake-up, briefly say babysitting is stopped, and end (see the babysit-stop skill). A wake-up-driven round is a continuation, never an explicit restart — when the turn input arrives prefixed as /loop /babysit-ci rather than a bare /babysit-ci, and/or you see a "resuming /loop wakeup" marker, it is a loop continuation that is subject to the stop check and must self-terminate if a stop exists in history. Do not treat it as the user re-starting, and do not let it "supersede" an earlier stop. Only a bare, user-typed /babysit-ci counts as an explicit resume.

3. If everything is green and complete, you are done — go to step 6.

4. If any check failed or was cancelled, fix and push:

   • Invoke the ci-failure skill with --fix to triage the failures and apply the fixes it is confident about.
   • If that produced working-tree changes, invoke the commit-commands:commit skill to commit them, then push.
   • If it produced no changes (the failure is a flake, infra/config, or pre-existing on main and not fixable in the repo), do not commit. Re-run the failed checks if they look like flakes; otherwise stop and report — the failure needs the user's judgment.

5. Repeat from step 2, watching the new run triggered by the push or rerun. Keep a round counter. After 5 rounds without reaching green, stop and report — something is not converging and needs the user.

6. Report back. Summarize: final CI state (green / still red / blocked on user), how many fix rounds it took, what was changed and pushed each round, and anything left for the user (flakes, infra, judgment calls).