CCAR-F scenarios › Multi-Agent Research System · Question 29 of 60
Scenario 3 · Multi-Agent Research System
You're building a research system on the Claude Agent SDK in which a coordinator agent delegates work to specialized subagents: one searches the web, one analyzes documents, one synthesizes findings, and one generates reports. Given a topic, the system produces a thorough report with citations.
Context Management & Reliability
The coordinator concatenates outputs from six subagents, about 60 pages, and passes them to the synthesis subagent. Final reports consistently cover findings from the first and last subagents well but often leave out important findings from the subagents in the middle of the input. What change best addresses this?
Show answer and explanation
Answer: A — Place a key-findings summary at the beginning of the aggregated input and organize the detailed results under explicit section headers for each subagent's contribution.
The lost-in-the-middle effect means content in the middle of long inputs is processed less reliably. A key-findings summary up front and explicit section headers help counter it (task 5.1). Randomizing order (b) only spreads the omissions around and makes the output unpredictable. An instruction to read carefully (c) doesn't change how position affects attention. Synthesizing each output separately (d) loses cross-source synthesis and produces a disjointed report.
More from this scenario
- Final reports often contain correct claims with missing or wrong citations. Traces show…
- The synthesis subagent's reports often include generic statements that appear in none of…
- For a broad topic, the coordinator delegates to four web search subagents covering…
- Users submit both broad questions ("How is climate policy affecting shipping costs?")…
- On the topic "remote work and urban real estate," the coordinator sends the same broad…