Tool interface design & boundaries
Design effective tool interfaces with clear descriptions and boundaries.
This subtopic (2.1) sits in Tool Design & MCP Integration (D2) on Anthropic's Claude Certified Architect — Foundations (CCA-F) exam. The bank holds 24 practice questions here — 6 easy, 11 medium, and 7 hard — with 5 free to try, answers and explanations included. 3 of the free questions are below; the rest are in the practice stream.
What the exam tests here
- tool descriptions as primary mechanism for LLM tool selection
- splitting generic tools into purpose-specific tools with defined contracts
- minimal descriptions cause misrouting among similar tools
- tool description content — input formats, example queries, edge cases, boundaries
Free practice questions: Tool interface design & boundaries
A customer support agent has two tools: lookup_order_by_id and lookup_order_by_email. Both descriptions say only 'Looks up an order.' The agent frequently calls the wrong tool, passing an email to lookup_order_by_id. What description change most directly fixes the routing failure?
Show answer & explanation
Correct answer: A. Add the accepted input type to each description: lookup_order_by_id accepts a numeric order ID; lookup_order_by_email accepts a customer email address
Tool descriptions are the primary signal the model uses for tool selection; adding the accepted input type directly to each description disambiguates the two tools at the selection step. 'System prompt rule' adds overhead and is probabilistic — the root cause is the minimal descriptions, not missing instructions. 'Renaming the tools' helps but provides less guidance than explicit input-type descriptions. 'Merging into one tool' eliminates the routing problem but changes the interface rather than fixing the description.
The coordinator agent in a research system must choose between search_web and search_arxiv for a user query. Both tools have no description. Which change would most directly help the agent select the right tool?
Show answer & explanation
Correct answer: A. Write a clear description for each tool explaining what it searches and when to use it
Tool descriptions are the primary mechanism the model uses to select among tools; adding clear descriptions directly enables correct routing between search_web and search_arxiv. 'Few-shot examples' in the system prompt can help but are secondary to the description itself. 'Renaming the tool' helps but a name alone cannot capture usage constraints as effectively as a description. 'One generic tool' removes the distinction entirely rather than enabling correct selection.
The coordinator agent must choose between the web search subagent and document analysis subagent for an incoming task. Both are registered tools. What is the primary mechanism Claude uses to decide which subagent to invoke?
Show answer & explanation
Correct answer: C. The tool description text provided for each subagent in the coordinator's tool definitions
Claude selects tools based on their description text — this is the primary signal for routing decisions. 'Order in which subagents are listed' has no semantic weight in tool selection. 'Response latency observed in prior turns' is not tracked or used by the model. 'Token count of each subagent's system prompt' is not a routing signal Claude has access to or uses.
2 more free questions on this subtopic in the practice stream, plus 19 in the full bank. Keep practicing →