Recursive Improvement Charter (Safety-First)

Draft a safety-first charter for recursive AI improvement: objectives, non-goals, guardrails, acceptable risk thresholds, and escalation policy. Include a one-page summary plus a detailed appendix.