loop-healthlisted

# /loop-health — lint diagnostic across all loop state ## What this checks Single pass, no mutation. Reports findings; user decides what to fix. ### A. Active cron loops (from `CronList`) For each: - **A1. Untracked in history** — no `created` / `resumed` / `discovered` event matches this cron's ID. Suggest `/loop-track <id>` or just call `/loops` (which auto-discovers as a side effect). - **A2. Missing canary** — prompt doesn't contain the "if you canceled or stopped this loop, you should NOT be seeing this message" pattern. Suggest `/loop-edit prompt-append: "Note: if you canceled or stopped this loop, you should NOT be seeing this message."`. - **A3. Oversized prompt** — prompt > 800 chars. Verbose prompts make each fire context-heavy; cleaving steady-state instructions into the prompt body wastes tokens on every fire. - **A4. Self-firing canary triggered** — if any recent history event shows this cron was `paused` or `stopped` AFTER its latest fire, but the cron still exists in CronList — the cancel didn't take. This is a real bug and warrants user attention. ### B. Paused loops (from `.paused-loops.json`) For each: - **B1. Stale paused** — `paused_at` is more than 5 days ago. The user probably forgot. Suggest `/loop-resume` or `/loop-stop`. - **B2. Missing canary** — same as A2. - **B3. Oversized prompt** — same as A3. - **B4. `inflight_tasks` orphaned** — entry has `inflight_tasks` but those tasks (by recorded ID) don't exist in current `Ta