What problem does it solve? Teams using persistent AI memory files (CLAUDE.md, MEMORY.md, topic files, dream logs) have no visibility into knowledge quality over time — which entries are stale, which are ready to promote, and how confidence decays across consolidation runs. This Skill turns those raw memory files into a single readable health report. ## Core Features & Use Cases - Memory health dashboard: Computes per-entry and aggregate metrics (confidence distribution, at-risk entries, promote candidates, file size budgets) and renders them as a self-contained memory/health.html file with inline SVG charts and no external dependencies. - Dream run history with justifications: Parses dream-log.md into a clickable operations table where each ADD/UPDATE/DELETE/PROMOTE row opens a side panel showing reason, evidence, before/after, and source session links. - Read-only and safe: Never modifies memory files; degrades gracefully to "No data yet" states when logs or topic files are missing. - Use Case: After several weeks of AI-assisted sessions, run the dashboard to see that 6 entries dropped below 0.3 confidence, 3 entries are ready to promote into CLAUDE.md, and MEMORY.md is approaching its 200-line budget. ## Quick Start Run the /dream-health command to generate the memory health dashboard at memory/health.html and open it in any browser.
