metrics

Example recommendations:

• "64% tier 1 work — your skill set matches well"
• "Adversary sample size < 5 — run /kernel:tearitapart more on tier 2+ work"
• "detect-secrets had 2 failures — check patterns are current"
• "3 learnings never reinforced in 30d — will be auto-pruned"

<on_start>

agentdb metrics
agentdb health

<ask_user> Use AskUserQuestion when: dashboard displayed and anomalies found Ask: "Anomaly in {metric}: {detail}. Want to investigate, or just note it?" Options: investigate, note and move on </ask_user> </on_start>