mxhealthlisted

# /mxHealth — Knowledge-DB Consistency Checker (AI-Steno: !=forbidden →=use ⚡=critical ?=ask) > **Context:** ALWAYS run as subagent(Agent-Tool) !main-context. Result: max 20 lines, problems only. Health-Check-Agent. Verify consistency of Knowledge-DB + local docs/. ## Trigger phrases This skill fires on: - `/health`, `/mxHealth` - Natural language: "health check", "check knowledge db", "verify consistency", "db health", "audit the project state" - Programmatic: pre-release validation, periodic `/loop` invocation, pre-commit integrity sweep ## MCP Required ⚡ mxHealth is MCP-dependent by design. Phases P1-P14 all query the Knowledge-DB. If `mx_ping` fails in Init step 2 → print `"MCP unreachable — /mxHealth requires MCP."` and ABORT. No partial runs, no local-only fallback mode. The caller should retry once MCP is back. ## Init 1. CLAUDE.md→`**Slug:**`=project. ∅slug→?user 2. mx_ping()→OK=continue | Error→"MCP unreachable — /mxHealth requires MCP." ABORT ## Phase 1: Load Inventory Execute in parallel: 1. `mx_briefing(project)` — Overview 2. `mx_search(project, doc_type='plan')` + `spec` + `decision` + `workflow_log` 3. Glob local: `docs/reference/*.md` 4. Read CLAUDE.md + docs/status.md 5. Count: DB-Docs total, local reference files, CLAUDE.md line count ## Phase 2: 14 Checks Per-check details (trigger, what is checked, severity, persistence target) → `Read ~/.claude/skills/mxHealth/references/checks.md.` | # | Check | Severity | |--