check-gh-actions

name	check-gh-actions
description	Use when user asks to check GitHub Actions for failing jobs, chronic cancellations, or jobs without a subsequent successful run. Also for CI/CD health audits, rerunning failed workflows, or monitoring workflow reliability. Aliases - checkGHA, gha_health, github_actions_health, gh_actions, check_actions, fixjob, github_actions, actions_health

Check GitHub Actions Health

Scan all recent GitHub Actions workflow runs on main, detect three classes of problems, and offer to fix them.

Three Failure Modes

1. Stale Failures (no subsequent success)

Workflows whose latest completed run on main is failure/timed_out/startup_failure/stale AND the workflow has NOT had a subsequent successful run since then.

Run this to find them:

gh run list --branch main --limit 200 --json workflowName,status,conclusion,databaseId,createdAt,url --jq '
  [group_by(.workflowName)[] |
   sort_by(.createdAt) | last |
   select(.status == "completed" and (.conclusion == "failure" or .conclusion == "timed_out" or .conclusion == "startup_failure" or .conclusion == "stale"))] |
  sort_by(.createdAt) | reverse
'

Show as a table, then ask to rerun. Use: gh run rerun --failed <run_id>

2. Chronic Cancellations

Workflows where the latest run is cancelled AND within the last 15 runs there are at least 4 cancellations with 0 successes and no success in the last 48 hours.

gh run list --branch main --limit 200 --json workflowName,status,conclusion,createdAt --jq '
  [group_by(.workflowName)[] |
   (. | sort_by(.createdAt)) as $runs |
   {wf: .[0].workflowName,
    latest: ($runs | last | .conclusion),
    cancelled: ([$runs[] | select(.conclusion == "cancelled")] | length),
    success: ([$runs[] | select(.conclusion == "success")] | length),
    total: ($runs | length)} |
   select(.latest == "cancelled" and .cancelled >= 4 and .total >= 5 and .success == 0) |
   {workflow: .wf, cancelled_count: .cancelled, total_runs: .total}
'

For each chronically-cancelled workflow, also check if there was any success in the last 48 hours:

# Replace WORKFLOW_NAME with actual workflow name
gh run list --branch main --limit 200 --json workflowName,status,conclusion,createdAt --jq '
  [group_by(.workflowName)[] |
   select(.[0].workflowName == WORKFLOW_NAME) |
   .[0].runs | [sort_by(.createdAt) | .[] | select(.conclusion == "success")] | last
'

If truly chronic (no recent success), retrigger with: gh workflow run NAME --ref main

3. Recurring Failures (same workflow failing multiple times in a row)

Check if any workflow has failed 3+ times consecutively without any intervening success:

gh run list --branch main --limit 100 --json workflowName,status,conclusion,createdAt --jq '
  [group_by(.workflowName)[] |
   (. | sort_by(.createdAt) | reverse | .[0:15]) as $recent |
   {wf: .[0].workflowName,
    fail_count: ([$recent[] | select(.conclusion == "failure")] | length),
    streak_starts: ($recent[0].conclusion)} |
   select(.fail_count >= 3) |
   {workflow: .wf, consecutive_failures: .fail_count, latest: .streak_starts}
'

Presentation

Show results as markdown tables:

STALE FAILURES (no next success): | Workflow | Conclusion | When | Run ID | Link |

CHRONIC CANCELLATIONS: | Workflow | Cancelled/Total | Recommended Action |

RECURRING FAILURES: | Workflow | Consecutive Failures | Latest | Action |

Fix Actions

• Stale failures: gh run rerun --failed <run_id>
• Chronic cancellations: gh workflow run NAME --ref main
• Recurring failures: gh run rerun --failed <run_id> plus investigate root cause

Context Gathering

Before rerunning, optionally run: gh run view <run_id> --log

Classify each failure as:

• FLAKY: intermittent, passes on retry
• BROKEN: always fails, needs code fix
• ENV: missing secrets (401/403), rate limits (429), external dependency down
• TIMEOUT: exceeded time limit

Notes

• The existing actions-failure-guardian.yml workflow does this automatically every hour via the GitHub API and posts to Discord. This skill is for on-demand CLI double-checks.
• Guardian script: scripts/actions_failure_guardian.py
• If gh auth fails, remind user to run gh auth login or check GIT_PAT_CLASSIC / GH_TOKEN env var.
• The --limit 200 covers about 3 days of hourly workflow runs. Adjust upward for longer scan windows.