syiibfs-hash/cc-agent-brain/cc-agent-brain-memory-audit

CC Agent Brain Memory Audit

Use when:

• many new events have accumulated
• the user says the assistant forgot something
• conflicting project information appears
• memory files are growing too large

Engine audit (primary)

python .claude/hooks/cc-agent-brain/brain_store.py audit
python .claude/hooks/cc-agent-brain/brain_store.py timeline --limit 20

The engine reports:

1. duplicate_active — same layer+title with several active copies (potential conflict)
2. broken_supersede_chain — a memory supersedes an id that no longer exists
3. stale_candidate — active semantic/working memory untouched for 90+ days
4. low_confidence — count of active low-confidence memories to verify
5. never_distilled — large event backlog with zero distilled memories

Markdown checks (secondary)

1. Duplicate preferences across memory files
2. Conflicting project status between engine and memory/active-projects.md
3. Important Wiki pages not reflected in memory
4. Sensitive information accidentally stored

Remediation

• Resolve duplicates by keeping one and superseding/expiring the others (memory expire <id>).
• Repair or relink broken chains.
• Re-confirm stale items with the user, then update or expire.
• Run sync-md after cleanup.

Output

Write a concise audit report to wiki/tasks/memory-audit-YYYY-MM-DD.md with issue ids and the action taken for each.