store-healthlisted

# store-health ## Purpose Single-call diagnostic audit of a Gramaton knowledge store. Bundles seven probes — total stats, pending classification, orphans, duplicates, stale temporal records, refuted/expiring records, heavily-connected concept candidates — into one structured report with actionable recommendations. The value over running `gramaton_stats` plus a few ad-hoc searches is the **cross-probe synthesis**: pending count interpreted against autonomous-curation availability ("pending > 0 AND autonomous: false → run curation-sweep"); orphan count interpreted against age ("many AGED orphans → capture pipeline isn't linking, investigate"); duplicate clusters interpreted against the auto-supersession threshold ("pre-existing pre-0.92 duplicates → optional dedup pass with consent"). Raw probe output gives you numbers; this skill gives you what to do. When NOT to use: - You only need one stat (use `gramaton_stats` directly). - You're mid-task and don't want a context-eating audit (run at task boundaries). - The store is brand new with under ~50 records (signals will be noisy and unactionable). When TO use: - User asks "is my store healthy?", "audit the store", or any general health question. - After a substantial ingest pass (benchmark extraction, bulk import) to confirm the store landed cleanly. - Periodically as a maintenance check on a long-lived store. Output is one structured report. Use MCP tools, not the CLI, unless MCP is unavailable. ## Sequence Run these in o