deploy-watch

name	deploy-watch
description	Monitor a deployment end-to-end after push. Finds the triggered CI run, polls until completion, downloads test artifacts, and reports results. Use after pushing to track the full deployment pipeline.
argument-hint	[repo]

Deploy Watch

End-to-end deployment monitoring after a push.

Steps

1. Detect repo: If $ARGUMENTS names a repo or subdirectory, use it. Otherwise detect from the current directory.

2. Find the triggered CI run:

   gh run list --limit 5 --json databaseId,status,conclusion,headBranch,workflowName,createdAt,event
   

   Pick the most recent run triggered by a push event. Show the run ID, workflow name, and branch.

3. Tell the user: "Watching run #ID (workflow) on branch. I'll check back periodically."

4. Poll for completion using iterative background checks (NOT a bash while loop - avoids 600s timeout):

   gh run view <run-id> --json status,conclusion
   

   Run this with run_in_background: true. When the notification comes back:

   • If status is still in_progress or queued: wait ~20 seconds, then run another background check
   • If status is completed: proceed to step 5
   • Repeat up to 30 times (covers ~10 minutes of deployment)

5. Check the result:

   gh run view <run-id> --json conclusion,jobs
   

   Parse jobs for pass/fail status and duration.

6. Download test artifacts if they exist:

   rm -rf /tmp/test-results
   gh run download <run-id> --name test-results --dir /tmp/test-results 2>/dev/null
   

   If artifacts exist, summarize test results (pass/fail counts, failed test names).

7. Report final status:

   • Run conclusion (success/failure)
   • Per-job status with duration
   • Test results summary (if artifacts existed)
   • If failed: gh run view <run-id> --log-failed | tail -50