Site Health Check
Runs a battery of drift and regression checks specific to one site — CTA routing, residual outdated prose, reference-data freshness, retired-model statuses, hub cross-linking and stray backup files under pages/.
Entry verified April 21, 2026
The short answer
Not a generic "is it up" monitor. Each site gets its own health skill (e.g. /pws-health, /bawt-health) encoding the drift risks that matter for that site — for PWS, course-push URLs outside the Tier A allowlist, stale claude-models.json data, and missing cross-links between the four Claude hub pages.
When to use it
Weekly as a portfolio sweep. After a major refactor or pivot. Before quarterly business reviews when regressions quietly accumulate.
Setup
- 1
Save the file as ~/.claude/commands/<site>-health.md (e.g. pws-health.md) — custom per site.
- 2
Each check is a tight grep or jq snippet: course URL outside Tier A, residual course prose, lastVerified age on data files, deprecated models past their retiresOn, hub pages missing cross-links, .bak/.orig/.backup files under pages/ or components/.
- 3
Findings are grouped by severity — HIGH (regression / stale / cruft), MEDIUM (prose / missing hub-link), LOW (data >14 days old).
- 4
Report-only by default — the skill doesn't fix anything unless asked.
Example
You: /pws-health Claude: [runs 6 checks] 3 findings: PROSE regression on best-ai-tools.js, HUB-LINK missing between 3 Claude hubs, data freshness 3d old (green).
Source & attribution
- Author
- Bryan Collins (pattern: /bawt-health, /pws-health, 2026-04-20)
- Licence
- MIT
- Type
- Original
Original pattern published under MIT — attribution preserved by convention, not licence requirement.
Caveats
Per-site bespoke — don't try to generalise into one skill. Every site has different drift.
56 skills across 10 categories, all licence-verified.