Goal stewardship — audit playbook

Use this when you need to explain what happened to a goal, or why stewardship behaved a certain way.

Sources of truth

1. Goal JSON — OPENCLAW_WORKSPACE + goalStewardship.goalsDir (default state/goals/<uuid>.json). Each file has history[] with timestamps, actors, and actions.
2. CLI
   • openclaw hybrid-mem goals list / status <id> — current state.
   • openclaw hybrid-mem goals audit — JSON snapshot (add --jsonl for one line per goal).
   • openclaw hybrid-mem goals stewardship-run — one deterministic health pass (watchdog logic).
3. ACTIVE-TASKS.md — If heartbeatRefreshActiveTask is on, the ## Active Goals section is a mirror refreshed on heartbeat; do not edit by hand.
4. Plugin logs — Look for memory-hybrid: goal stewardship and ACTIVE-TASKS.md mirror refreshed.
5. Event log (when configured) — May contain goal.* events from tools and watchdog.

Typical questions

Question	Where to look
Why was a goal blocked?	status + currentBlockers + history in goal JSON; health check actions from stewardship-run.
Why no stewardship this heartbeat?	Last user message must match a heartbeat pattern (defaults include “heartbeat”, “scheduled ping”). Check goalStewardship.heartbeatPatterns.
Why only some goals in the prepend?	Multi-goal cap (multiGoalMaxChars, multiGoalMaxGoals), cooldown per goal, weights (attentionWeights), round-robin cursor file _stewardship_rr.json in goals dir.
Why confirm on register?	confirmationPolicy.requireRegisterAckForPriorities — use confirmed: true on goal_register after user approval.

Automation

Pipe goals audit --jsonl into your log stack or CI artifact store for long-term correlation with gateway logs.