| name | site-health |
| description | Run the "anything broken?" sweep for the LMS — localhost Docker stack and/or production HTTP — and get a PASS/FAIL table. Use when asked to "check site health", "is anything broken", "smoke test", "verify localhost before push", after container restarts, after theme/CSS changes, or as the mandatory verification gate before the admin decides to push. Delegates to the site-health-checker agent. |
Site Health Check
- Determine scope from the request:
localhost, prod, or both (default: both if Docker is up, else prod-only).
- Launch the site-health-checker agent (Agent tool,
subagent_type: site-health-checker) with the scope and, if the check follows a specific change, the list of changed files so it weights the affected routes.
- Relay the agent's PASS/FAIL table to the user verbatim, then add a one-paragraph verdict: healthy / degraded (what + why) / broken (what to fix first).
- If the check was requested as a pre-push gate and everything passes, remind the user: pushing to
main deploys to production via Coolify — the ADMIN decides whether to push; do not push automatically. If anything fails, the push gate stays closed.
- For failures, route the fix: styling → backend-design skill; caching/speed → caching-speed-optimizer agent; DB → mysql-memory-optimizer agent; security header/exposure → openmage-security-auditor agent; broken route/fatal → fix inline with the pre-push checklist from CLAUDE.md.
Frontend invariant: any change made in this repo must keep the customer-facing storefront looking IDENTICAL. If the health check follows storefront-adjacent edits (skin/frontend/, app/design/frontend/, layout XML), additionally screenshot the homepage + one course page with Playwright at 1440px and 390px widths and compare against var/health-baseline/ screenshots (create them on first run) — flag any visual diff.