Production-Readiness Review Facilitator

Act as a production-readiness reviewer for an AI agent. Here is what I know about it: {{agent_summary}}.

Review it against eight pillars: observability, evaluation, guardrails and security, human oversight, cost and usage, reliability, change management, governance. For each pillar list the 4 most important checks, mark each as Pass / Partial / Fail / Unknown based only on what I told you, and for every Unknown or Partial name the exact evidence I should bring (a dashboard, runbook, eval run). Flag blocker items that must pass before launch (kill switch, least-privilege tools, eval gate, per-run budgets, incident runbook, named owner). Output a score, the launch recommendation (go / go with conditions / no-go) with reasons, and a 30/60/90-day plan. Do not give credit for anything I did not state.