memory-index-compact

# Memory Index Compact ## Purpose / When to Activate Activate when: - A registry file (`index.md`, `index-*.md`, archive index) breaches its file-size budget (typically 40k chars). The `post-commit` hook surfaces this. - An index row has accreted substance that duplicates the target file body (Tier 2 cycle trails, commit SHAs, §-numbered substance dumps, validation ceremony, drift-heal forensic prose). Detection : any row > 200 chars in a registry table is suspect. - Scheduled hygiene pass during `memory-index-sync` or post-DEC-activation cleanup. **Distinct from other memory skills:** - `memory-compact` — compacts memory CONTENT (sessions, domains) into summaries. NOT for index files. - `memory-archive-superseded` — migrates superseded DEC rows to archive. Adjacent but separate. - `memory-index-sync` — heals drift between filesystem and registry (missing rows). Detects bloat ; compaction is THIS skill's job. ## Core principle — index = pointer Every row in a registry must answer one question : *« Does an agent reading this row for relevance assessment need this word? »* - **Keep** : file path, ID, category, level, status, ≤30-word topic essence, relation to other entities (amends, supersedes, related). - **Dégage** : Tier 2 cycle trails, finding counts (F-C1, REFINE 2C+7H+4M+4L), commit SHAs, validation ceremony ("Fred ratified post Tier 2 + Compliance 11/11 PASS"), §-numbered substance dumps duplicating DEC body, drift-heal forensic prose, self-referential pointers wi...