CCARF Practice Question

Paper 4 · Question 15

D1 Agentic Architecture & Orchestration · Paper 4

In a hub-and-spoke research system, a coordinator manages several search and analysis subagents. Which responsibilities sit with the coordinator rather than with individual subagents? (Select 3)

• A.Deciding which subagents to invoke based on the complexity of each incoming query.
• B.Running every search itself first so subagents only verify its findings.
• C.Routing all inter-agent communication so subagents never exchange messages directly.
• D.Sharing its full conversation history with each subagent automatically at spawn time.
• E.Aggregating subagent results and re-delegating targeted follow-up queries when synthesis shows gaps.

Show answer & explanations

Correct answer: A, C, E

A: The guide places dynamic subagent selection with the coordinator, which analyses query requirements instead of always running the full pipeline.

B: The coordinator delegates work rather than duplicating it; doing every search itself defeats the point of decomposition.

C: Hub-and-spoke routes every message through the coordinator for observability and consistent error handling.

D: Subagents run with isolated context; nothing is inherited automatically, and the coordinator passes only the context each task needs.

E: Result aggregation and gap-driven re-delegation are coordinator responsibilities in the iterative refinement loop.

Related Agentic Architecture & Orchestration questions

More: Paper 4 · All practice questions