| name | helm-health |
| description | Report on Helm routing health — escalation rate, learned patterns, routing efficiency. Use when asked about engine routing performance or during sitrep generation. |
| version | 1.0.0 |
| author | quartermaster |
| tags | ["helm","routing","metrics","efficiency"] |
| intent | Efficient [I06], Observable [I04] |
helm-health
Report on the Helm (engine-router) routing system health.
When to use
- During sitrep generation (include Helm metrics)
- When Reactor asks about routing efficiency
- Weekly review of engine routing patterns
- After significant changes to engine configs
Process
- Read
/root/.openclaw/logs/escalation.log — count escalations and de-escalation hints
- Read
/root/.openclaw/engines/helm-learned.json — count learned patterns
- Read
/root/.openclaw/logs/helm-metrics.json — check trend (escalation rate over time)
- Compute:
- Current escalation rate (escalations / total tasks routed)
- Week-over-week trend (improving, flat, degrading)
- Top unresolved bounce reasons (patterns not yet learned)
- Cost of escalation waste (sum of wasted tokens from misroutes)
- Report concisely:
- Escalation rate: X% (trend: ↓ improving / → flat / ↑ degrading)
- Learned patterns: N (last learned: date)
- Top bounces: list of frequent escalation reasons not yet covered
- Recommendation: "run helm-learn --apply" if new patterns pending
Output Format
Include in sitrep under "## Helm Routing" section. Keep under 10 lines.