loop-on-ci

To fix failing GitHub pull request checks, you can loop on CI by using `gh pr checks` as the source of truth, prioritizing already-failing checks over pending ones, and re-validating the entire PR check set after each push until all required checks are green.

You use `gh pr checks` to monitor CI failures by inspecting its JSON output to track the full set of PR-attached checks, using watch mode with fail-fast to reduce wait time, and pulling logs from linked GitHub Actions runs to apply targeted fixes.

The best way to debug CI failures without guessing is to systematically diagnose the current PR state, pull real failure evidence from the linked GitHub Actions run using `gh run view --log-failed`, and apply scoped fixes before re-checking the PR.

Yes, this CI debugging approach works with non-GitHub Actions checks because it uses `gh pr checks` as the source of truth for all checks tied to the pull request, including non-GitHub Actions signals, rather than relying solely on GitHub Actions run data.

To prioritize which CI checks to fix first, you diagnose the current PR state and immediately address already-failing checks before waiting on pending ones, using watch mode with fail-fast to reduce overall wait time during the debugging loop.

A limitation of using `gh pr checks` for PR debugging is that you must enforce guardrails to keep fixes scoped and avoid bypassing verification, and while it tracks all PR-attached checks, pulling detailed failure logs is only applicable when checks are linked to GitHub Actions runs.