claws-statslisted

## What this is `/claws-stats` is the Claws observability dashboard. It reads `.claws/events.log` and the live MCP bus to render a snapshot of terminal state, active workers, wave phase, memory recall counts, estimated token spend, and mission delivery health. All reads are on-demand; no background process is created. Flags: `--delivery` / `--concurrency` / `--context` / `--json` / `--memory-only` / `--wave-only` / `--cu-only` / `--watch` (any panel). ## When to invoke - After a wave dispatch — verify workers received missions (`--delivery`). - Large fleet (N ≥ 6) — check adaptive ceiling and queue depth (`--concurrency`). - Workers appear memory-starved — inspect slug counts and budget (`--context`). - Incident triage — identify failure cause by runbook ID. - After tuning `CLAWS_POOL_SIZE` / `CLAWS_CONCURRENCY_CAP` — confirm impact. ## When NOT to invoke - To fix problems — use [[claws-fix]] for targeted repair instead. - As a substitute for live bus monitoring — arm `Monitor` with `stream-events.js` for real-time streams. - For memory search or recall — use [[claws-memory]] tools. - During sub-worker missions — workers should not self-introspect via stats. ## Steps 1. Determine which panel you need: delivery, concurrency, context, or overview. 2. Run `/claws-stats [--flag] [--json] [--watch]`. 3. **`--delivery`**: read success rate, first-ack latency p50/p95, and failure-mode breakdown by runbook ID. 4. **`--concurrency`**: inspect pool state, queue depth, in-flight