最も有効な CCAR-F 試験参考書

During testing, when a customer says, "I need a refund for my recent purchase," the agent immediately invokes process_refund but populates the required order_id parameter with a plausible-looking fabricated value instead of first calling lookup_order. The refund fails because the invented order identifier does not exist. Which change directly addresses the root cause of the fabricated order_id?

• A. Preprocess customer messages to extract any mentioned order identifiers and inject them into the conversation before sending the request to Claude.
• B. Change tool_choice from auto to any so Claude must call a tool on every turn.
• C. Update the process_refund tool description to state explicitly that order_id must come from a successful lookup_order result and must never be assumed, inferred, or invented.
• D. Add server-side validation that checks whether order_id exists before attempting the refund and returns an error when it does not.

The document analysis agent has a single analyze_documnet tool that takes a document and a free-text instruction parameter. During evaluation, requests like "extract the key financial metrics" often return narrative summaries, while "summarize the methodology" sometimes returns raw data tables. The synthesis agent reports that 35% of analysis results require re-requests with clarified instructions. What's the most effective way to improve reliability?

• A. Enhance the tool description with detailed examples showing how different instruction phrasings should map to different output formats
• B. Keep the single tool but add an analyze_type enum parameter requiring explicit selection between extraction, summarization, and verification modes
• C. Have the coordinator pre-classify each analysis request before passing instructions to the document analysis agent
• D. Split the generic tool into purpose-specific tools - extract_data_points, summarize_content, verify_claim_against_source - each with defined input/output contracts

The coordinator agent has AgentDefinitions configured for all four specialized subagents, each with appropriate descriptions, prompts, and tool restrictions. During testing, you notice the coordinator correctly reasons about when to delegate - it generates messages like "I'll ask the web search agent to find sources on this topic" - but no subagent execution ever occurs. The coordinator then proceeds as if the delegation happened and continues with incomplete information. Logs show no errors. What is the most likely cause?

• A. The coordinator's max_tokens setting is too low, causing the Task tool invocation to be truncated before the subagent type parameter can be specified.
• B. The AgentDefinitions are configured correctly, but the coordinator's system prompt doesn't explicitly list the available subagent types, preventing the model from knowing they can be invoked.
• C. The coordinator's allowedTools configuration doesn't include "Task", so while it can reason about delegation, it cannot invoke the tool required to spawn subagents.
• D. Subagent context isolation means task descriptions from the coordinator don't automatically reach subagents; you need to configure explicit context forwarding in ClaudeAgentOptions.

You are integrating Claude Code into your Continuous Integration/Continuous Deployment (CI/CD) pipeline. The system runs automated code reviews, generates test cases, and provides feedback on pull requests. You need to design prompts that provide actionable feedback and minimize false positives.
Your automated code review is missing genuine bugs in pull requests. Investigation reveals that your review prompt includes the instruction: "Only flag critical issues that would definitely cause production failures. Ignore minor concerns and anything you are uncertain about." Developers confirm that some missed bugs are genuine logic errors that the model investigated but chose not to report. The team requires the review output to remain structured, with each finding tagged with metadata, and actionable.
Which prompt change both removes the cause of the suppressed findings and preserves structured, tagged output for downstream filtering?

• A. Instruct the model to report all findings with confidence and severity tags, deferring filtering to a downstream step.
• B. Add a second review pass that rereads the diff using the same prompt, looking for anything the first pass may have missed.
• C. Remove all severity-related instructions from the prompt and let the model use its default judgment about what to report.
• D. Enable extended thinking and instruct the model to reason step by step about every code change before producing its review.

A customer contacts the agent about a warranty claim on a power drill. Resolving this requires multiple sequential tool calls: get_customer to look up their account, lookup_order to find the purchase details, and then either process_refund or escalate_to_human depending on warranty eligibility. You're implementing the agentic loop that orchestrates these steps using the Claude API. What is the primary mechanism your application uses to determine whether to continue the loop or stop?

• A. You track the number of tool calls made and exit the loop once a preconfigured maximum is reached.
• B. You check whether Claude's response contains a text content block - if text is present, the agent has produced its final answer and the loop should exit.
• C. You manually set the tool_choice parameter to "none" after the final expected tool call to force Claude to stop requesting tools.
• D. You check the stop_reason field in each API response - the loop continues while it equals
  "tool_use" and exits when it changes to "end_turn" or another terminal value.