AgentSkillsCN

Integrity Scan

完整性扫描

SKILL.md

/scan — Memory Integrity Scan

Run the intelligence scanner to detect contradictions, drift, dead decisions, and missing cross-references across the entire memory workspace.

When to Use

  • •After a session with significant decisions or changes
  • •During weekly maintenance
  • •When something "feels wrong" about the memory state
  • •Before making architectural decisions (verify no contradictions)

How to Run

bash
python3 maintenance/intel_scan.py "${MIND_MEM_WORKSPACE:-.}"

What It Checks

  1. •Contradictions — Decisions that conflict with each other (via ConstraintSignatures)
  2. •Drift — Decisions referenced in daily logs but never formalized
  3. •Dead decisions — Active decisions with no recent references
  4. •Orphan tasks — Tasks referencing non-existent decisions
  5. •Impact graph — Downstream effects of each decision
  6. •Coverage — How many files are monitored vs total

Output

  • •Updates intelligence/CONTRADICTIONS.md, intelligence/DRIFT.md, intelligence/IMPACT.md
  • •Updates memory/intel-state.json with latest metrics and counters
  • •Saves state snapshot in intelligence/state/snapshots/
  • •Generates intelligence/BRIEFINGS.md weekly briefing

Modes

  • •detect_only — Report issues, never propose changes (default)
  • •propose — Report issues AND generate fix proposals in intelligence/proposed/

Check current mode: cat memory/intel-state.json | python3 -c "import json,sys;print(json.load(sys.stdin)['governance_mode'])"