PREPARING FOR CERTIFICATION

MAPPING THE LAYERS TO STUDY DOMAINS A study domain is a group of related topics that you review together. The seven layers give a natural backbone, and the other parts of the book fit around them. Here is a suggested map. Domain 1, Foundations and the Map. Parts One and Two, Chapters 1 through 8. You should be able to explain what an agent and a harness are, why harness engineering emerged, and what the seven ETCLOVG layers are. Domain 2, Running and Controlling Agent Runs. Part Three on Execution, Chapters 9 through 12, and Part Six on Lifecycle, Chapters 23 through 26. This domain covers the agent loop, state, routing, starting and ending runs, hooks, and budgets. Domain 3, Giving the Agent Capabilities. Part Four on Tooling, Chapters 13 through 17, and the ecosystem material in Part Twelve. This covers tools, registries, good tool design, the MCP, frameworks, and standards. Domain 4, Managing Working Memory. Part Five on Context, Chapters 18 through 22, and Part Eleven on memory and retrieval. This covers context rot and bloat, the context builder, compaction, memory tiers, and RAG. Domain 5, Measuring and Understanding Behavior. Part Seven on Observability, Chapters 27 through 30, and Part Eight on Verification, Chapters 31 through 35. This covers traces, spans, logs, metrics, replay, evaluation design, scoring, and testing the harness itself. Domain 6, Governing Safety and Risk. Part Nine, Chapters 36 through 41. This covers safety architecture, permissions, prompt injection, tool poisoning, supply-chain security, isolation, and responsible use. Domain 7, Scaling and Operating Real Systems. Parts Ten, Thirteen, Fourteen, and Fifteen. This covers multi-agent patterns, cost, latency, reliability, deployment, Agent Operations (AgentOps), incident response, and specialized harnesses. The profession chapters, Part Seventeen, and the Practice Lab, Part Sixteen, support every domain, because the lab gives you hands-on evidence for each layer.