[internal-skills] fetch CI status on demand instead of blocking on --wait (#101661)
The `$pr-status-triage` skill prescribed `node scripts/pr-status.js [PR]
--wait` as the default first step for CI triage, which blocks on `gh run
watch` until the entire workflow completes before any triage begins. In
practice this left the agent idle waiting for full workflow completion
instead of triaging failures as they landed.
This updates the skill to fetch current status on demand (no `--wait`)
and start triaging immediately from the jobs that have already finished,
re-fetching as more results land. `--wait` is now reserved for cases
where the final run state itself is needed: confirming an all-green run
before handing back, or picking up results after retriggering jobs. The
flaky-rerun loop uses a `--wait` re-analysis instead of a flat 5-minute
sleep, and a new step makes push timing explicit: push once every result
relevant to the fix is in, without waiting for the whole workflow to
complete.
<!-- NEXT_JS_LLM -->
---------
Co-authored-by: Claude Code (Kimi K3) <noreply@anthropic.com>