CI Health Check

Scan CI status across repos in an org. Run from ~/code/{org}/ or pass the org name as an argument.

Helper script

A pre-built script at ~/.claude/scripts/org-ci-health.sh collects all CI data in one pass. Run it first to avoid spending tokens on individual gh calls.

~/.claude/scripts/org-ci-health.sh <org-name> > /tmp/org-ci-health.json

If the script is missing or not executable, flag this to the user and stop.

The script outputs a JSON array with per-repo entries containing:

repo, gh_org, has_workflows
recent_runs (last 5 CI runs on main with status/conclusion)
actions_used (list of uses: action references from workflows)

Phase 1: Analyse script output

Parse the JSON and flag:

Failing: latest main CI conclusion is not "success"
Stale: no CI run in 30+ days
No CI: has_workflows is false

Phase 2: Check workflow versions

From actions_used, flag:

Actions using deprecated versions (e.g. actions/checkout@v3 when v4 exists)
Actions pinned to branches instead of tags
Outdated actions/setup-r or julia-actions/setup-julia

Phase 3: Check for common issues

Workflows using deprecated set-output or save-state commands
Missing permissions: blocks
Workflows that run on push to all branches (should be limited)
R packages missing R-CMD-check workflow
Julia packages missing CI workflow

Phase 4: Report

Present a dashboard:

Repo	Main CI	Last run	Actions up to date	Issues

Then list specific issues grouped by repo.

Phase 5: Fix (with confirmation)

For fixable issues (action version bumps, workflow updates):

Create a worktree branch
Apply the fix
Create a PR

Use a team if more than 3 repos need fixes.

Auto-Exit When Standalone

IMPORTANT: If this command is being run as a standalone request, automatically exit after completing all phases successfully.