graph-hygiene

# Graph Hygiene. Logseq Knowledge Graph Maintenance ## Tone & Voice Diagnostic operational telemetry. Output is structured findings + actionable severity tags, not advisory prose. Match the self-audit register: terse, evidence-grounded, severity classified. **Hard rule**: never auto-delete pages. flag for `/review` always, deletion is human-approved only. ## Invocation patterns How to invoke: - Scheduled cron: `graph-hygiene-weekly` fires weekly (e.g. Sat 08:00 local) → invokes this skill autonomously - Manual: `/review graph`, "run graph hygiene", or "graph health check" in chat - Trigger keywords: graph, hygiene, lint logseq, orphan check, stale pages ## Purpose Automated lint and health check for the Logseq knowledge graph. Detects entropy before it compounds: stale pages, broken links, orphans, missing properties, inconsistent tags. ## When to Run - Scheduled: weekly (e.g. Sunday 08:00 local) - Manual: `/review graph` or when the operator asks about graph health ## Checks ### 1. Stale Pages. `severity: critical` (Pipeline/Client) / `warning` (Prospect) Pages in active namespaces (Pipeline/, Client/, Prospect/) where `updated::` property is older than 30 days. - **Action:** List with last-updated date. Flag as "needs review or archive." - **Threshold:** Pipeline/Client >30d = critical; Prospect >60d = warning. ### 2. Orphan Pages. `severity: warning` Pages with zero inbound links (no other page references them). - **Exclude:** Journal/, Template/, Meta/, Archive/ ...