Cloud native agentic standards

CNCF's foundational standards document for deploying agentic AI systems in cloud-native (Kubernetes) environments. Covers four key areas: general container best practices (security, observability, fault tolerance), control and communication (MCP, A2A, gRPC, Kafka, agent registries), observability (MELT stack, OpenTelemetry traces/metrics for LLM inference), governance (pre-deployment evaluation, synthetic testing, explainability, LLMOps lifecycle), and security (agent identity with SPIFFE/SPIRE, zero-trust tenancy, data access controls, prompt injection mitigation). Emphasizes that standardization is still nascent and the document will evolve rapidly as protocols like MCP and A2A mature.