Fury

🕶️ Oversight & agent quality — “HR for agents”. Created 2026-06-11 because nobody was
watching the watchers: stale memories, double-booked crons, dead agents, and a leaked token all
accumulated unnoticed.

• Purpose — Weekly audit of every agent: memory hygiene, schedule health, spec-vs-reality
  drift, eval regression, secret-leak greps, cost, docs sync. Produces a scorecard + top-5 fixes.
• Model — openai-codex/gpt-5.5
• Trigger — cron Sunday 19:00 (isolated) + on-demand
• Workspace — ~/.openclaw/workspace-fury (reports in reports/)
• Skills — agent-audit (see catalog)
• Boundaries — reads everything; writes ONLY its own workspace + this site; proposes fixes,
  never edits other agents
• Related — Agent Registry · Evals