GHA Checks

Quickly check GitHub Actions CI status for a PR or branch, summarize results, and triage failures.

When to Use

• User asks about CI/check status on a PR or branch
• User wants to know if checks passed before merging
• User asks to re-run failed jobs
• After pushing, to confirm CI is green

Process

Step 1: Resolve the Target

Determine what to check:

• PR number given (e.g., "checks on #319"): use that directly
• PR URL given: extract owner/repo and PR number
• Run URL given (e.g., github.com/.../actions/runs/123): use /gha-log-reader instead — it's built for run-level diagnosis
• No PR specified: find the PR for the current branch:

  gh pr view --json number,url,headRefName --jq '"\(.number)|\(.url)|\(.headRefName)"'

• No PR exists: fall back to listing runs for the branch:

  branch=$(git branch --show-current)
  gh run list --branch "$branch" --limit 5 --json databaseId,status,conclusion,name,createdAt,headBranch \
    --jq '.[] | "\(.databaseId)|\(.name)|\(.status)|\(.conclusion // "—")|\(.createdAt)"'

Step 2: Get Check Status

For a PR:

gh pr checks <number> --json name,state,description,detailsUrl,completedAt \
  --jq '.[] | "\(.name)|\(.state)|\(.description // "")|\(.detailsUrl)"'

Step 3: Summarize

Present a concise summary:

## PR #<number> Checks

✅ 3 passed | ❌ 1 failed | ⏳ 1 pending

| Check | Status | Details |
|-------|--------|---------|
| build | ✅ pass | — |
| lint  | ✅ pass | — |
| test  | ❌ fail | Node 20 — exit code 1 |
| deploy | ⏳ pending | Waiting on test |

Rules:

• Show all checks, not just failures
• Sort: failed first, then pending, then passed
• Include the description field when available (often has the failure hint)
• Keep it scannable — one line per check

Step 4: Triage Failures

For each failed check:

1. Get the run ID:

   gh pr checks <number> --json name,state,detailsUrl \
     --jq '.[] | select(.state == "FAILURE") | "\(.name)|\(.detailsUrl)"'

2. Extract the run ID from the URL (format: .../actions/runs/<id>/...)

3. Get the failed step and tail of its log:

   gh run view <run-id> --json jobs \
     --jq '.jobs[] | select(.conclusion == "failure") | "\(.name): step \(.steps[] | select(.conclusion == "failure") | .name)"'

4. Pull the last 50 lines of the failed job's log:

   gh run view <run-id> --log-failed 2>/dev/null | tail -50

5. Present a brief diagnosis:

   ### ❌ test (run 12345678)
   Failed step: Run tests
   
   <last few relevant log lines>
   
   Likely cause: <one-sentence assessment>

Step 5: Offer Next Actions

Based on the results:

• All green: "All checks passed. Ready to merge."
• Failures: Offer:
  • "Want me to dig deeper into the <name> failure?" → invoke /gha-log-reader with the run URL
  • "Want me to re-run the failed jobs?" → gh run rerun <run-id> --failed
• Pending: "Checks still running. Want me to check again in a few minutes?" → suggest /loop
• No checks: "No checks configured for this PR/branch."

Re-running Jobs

When the user asks to re-run:

# Re-run only failed jobs (cheaper, faster)
gh run rerun <run-id> --failed

# Re-run all jobs in the run
gh run rerun <run-id>

Always prefer --failed unless the user explicitly asks to re-run everything.

Integration

• Escalate to /gha-log-reader for deep log analysis when the tail snippet isn't enough
• Pairs with /git:push — check status after pushing
• Pairs with pr-style — confirm checks before marking ready for review

Do NOT

• Parse full logs inline — that's /gha-log-reader's job
• Re-run jobs without asking the user first
• Assume a failed check is the user's fault — it could be infra, flaky tests, or upstream