Test yourself on this chapter. Try each question before reading the answer.
1. When does an agent make sense?
- A. Any time you want a fancy prompt
- B. When the task is multi-step, observable, tool-supported and verifiable
- C. Only for writing emails
- D. When you do not know the goal
Answer: B. Agents need a goal, environment, actions, observations and verification. Otherwise a workflow or chat is usually better.
2. When are subagents useful?
- A. When work can be split into independent streams
- B. When every task has one paragraph
- C. When no synthesis is needed
- D. When you want to avoid verification
Answer: A. Subagents are useful for separable work that can be synthesized by the main agent.