weekly-site-health-auditlisted

# Weekly site health audit Runs [`site-health-audit`](../site-health-audit/) on a weekly cadence with one addition: a week-over-week diff. Use the config at `../site-health-audit/config.yml` (copy it from `config.example.yml` there if you have not yet). The audit's run frame, state file, evidence rule, and report template all apply unchanged — this skill adds nothing to them. ## The diff Compare against the newest dated report in the config's `report.dir`. For every tracked metric, report the previous value, the current value, and the direction of change. When no previous report exists, say so plainly in the report header and record every value as the new baseline. Do not silently present a first run as though it were a comparison — a baseline labelled as a trend is worse than no trend at all. ## Why weekly Site health moves on a scale of days, not hours. A daily run produces noise you learn to ignore, and an ignored report is the same as no report. Weekly is frequent enough to catch a regression while you can still connect it to what changed, and rare enough that you actually read it. Schedule it with whatever your platform provides — cron, launchd, systemd timers, or your agent's own scheduled tasks. Pick a time you will read the output.