- 设计与结构主轴：OpenAI
A practical guide to building agents/Agents SDK Overview/Context management/Sessions、AnthropicBuilding Effective AI Agents - 评测与治理主轴：OpenAI
Evaluating model performance/Safety in building agents、MicrosoftTracing - 组织与风险主轴：
NIST AI RMF 1.0、NIST GenAI Profile - 权限与运行主轴：MCP
Authorization、OpenAILatency optimization、The Tail at Scale、Circuit Breaker - 本章落点主轴：成熟的 Agent 架构必须能在目标、对象、模型、状态、动作、控制流、评测、治理、运行和演化十个面上逐项自圆其说
附录 G：分章资料清单