Securing AI Agents in the Enterprise: NanoClaw, Zero Trust Guardrails, and Governance at Scale