pr-watch-fix

Watch the open PR for the current branch, wait for CI, and fix failures.

Delegation hints

• To gather CI status and failure excerpts cheaply, delegate the polling/log-fetch step to the pr-monitor agent (haiku). Use it instead of running gh pr checks + log-fetch loops directly from this skill. Reserve this skill's main loop (sonnet) for deciding fixes, editing code, and pushing.
• When a failure matches a recognized class, hand the diagnosis-and-fix off to a specialist skill instead of solving it inline — see Step 5.1 — Route recognized failures to a specialist skill. This keeps CVE reasoning, test-fixture debugging, and issue-scoped fixes in the skill that knows them best.

Loop

Repeat until the PR is fully green or you stop intentionally:

0. Stop any prior PR-watch Monitor

Before doing anything else, cancel a previous run's still-running poller. Re-invoking /pr-watch-fix always wins — the new run resets state.

Use TaskList to find Monitors whose description starts with PR watch: (the convention used by step 3) and call TaskStop on each. Do not stop tasks that don't start with this prefix — they belong to other work.

1. Find the PR