CCARF Practice Question

Paper 3 · Question 3

D1 Agentic Architecture & Orchestration · Paper 3

[Scenario 1: Customer Support Resolution Agent] You are building a customer support resolution agent with the Claude Agent SDK. It handles high-ambiguity requests (returns, billing disputes, account issues) via custom MCP tools — get_customer, lookup_order, process_refund, escalate_to_human — targeting 80%+ first-contact resolution while knowing when to escalate. Company policy caps autonomous refunds at $500; larger refunds must go to a human. In testing, prompt instructions alone still let an occasional $650 refund through. What is the correct enforcement design?

• A.Implement a hook that intercepts outgoing process_refund calls, blocks any amount above $500, and redirects the agent to the escalate_to_human workflow
• B.Move the $500 limit to the top of the system prompt and repeat it in every user turn
• C.Lower the model temperature so it follows the policy more consistently
• D.Add the limit to the process_refund tool description so the model reads it during selection

Show answer & explanations

Correct answer: A

A: Business rules with financial consequences need deterministic guarantees. A tool-call interception hook enforces the threshold in code — the violating call can never execute — while prompt-based approaches remain probabilistic.

B: and D improve the odds but retain a non-zero failure rate, which the scenario shows is unacceptable here

C: misapplies a sampling parameter to a compliance problem

D: This option does not best satisfy the scenario requirements.

Related Agentic Architecture & Orchestration questions

More: Paper 3 · All practice questions