INCIDENT RESPONSE FOR AGENTS

CHECK YOUR UNDERSTANDING • What is triage for? ◦ Answer: Ranking incidents by severity so the most urgent are handled first. • Why contain before finding the root cause? ◦ Answer: Stopping ongoing harm is the top priority; analysis can continue afterward. • Why must postmortems be blameless? ◦ Answer: Blame makes people hide mistakes, while a system focus encourages honesty and finds structural fixes. • How does an incident become a regression test? ◦ Answer: The exact failing situation is added to the evaluation suite so every future release is checked against it. • What makes an action item useful? ◦ Answer: A specific task with a named owner and a due date that is followed up. PART FIFTEEN: SPECIALIZED AND APPLIED HARNESSES Everything you have learned so far describes the general ideas that apply to nearly every harness. Part Fifteen shows how those ideas bend and stretch when a harness is built for one particular kind of work. Chapter 61 looks at coding agents, which are among the most widely used agents today. Chapter 62 looks at agents that operate a computer screen or a web browser. Chapter 63 looks at agents that research, write, and manage knowledge. Chapter 64 looks at domain-specific harnesses, such as customer support, data analysis, regulated industries, and agents that run on small local devices. As you read, notice that the layers do not change. Every one of these harnesses still needs an execution loop, tools, context, a lifecycle, observability, verification, and governance. What changes is which layer carries the most weight.