You must deterministically block any refund transaction exceeding $500 and redirect it to a human supervisor. How should you implement this using the Claude Agent SDK?

• A.Implement a tool call interception hook (e.g., PreToolUse or PostToolUse) that programmatically blocks the policy-violating action and redirects to an escalation workflow.
• B.Provide 5 negative few-shot examples in the prompt demonstrating the agent refusing high-value refunds.
• C.Include a capitalized, bold warning in the agent's system prompt forbidding refunds over $500.
• D.Use a Notification hook to alert a manager while simultaneously allowing the transaction to proceed.

Comment: *

Name: *

Email: *

Verification: *

When formatting few-shot examples, what is the most important structural rule to follow to ensure the model produces the desired output?

• A.Format consistency: all examples must follow the identical output structure.
• B.Randomization: examples should be presented in different formats to increase model adaptability.
• C.Conciseness: output examples should be truncated to save tokens.
• D.Exclusion: never use XML tags around the examples.

Comment: *

Name: *

Email: *

Verification: *

Why is it heavily recommended to use Agent SDK hooks rather than prompt instructions to enforce critical business rules in a application?

• A.Hooks inherently consume zero tokens from the context window.
• B.Hooks provide deterministic guarantees, whereas prompt instructions rely on probabilistic compliance and carry a non-zero failure rate.
• C.Hooks automatically implement strict JSON schema validation for all tool_use calls.
• D.Prompt instructions cannot be modified without restarting the entire MCP server architecture.

Comment: *

Name: *

Email: *

Verification: *

A pull request modifies 14 files across a stock tracking module. When analyzing all files in a single pass, Claude provides superficial comments, misses obvious bugs, and produces contradictory feedback. How should you redesign the multi-pass review architecture to resolve this?

• A.Switch to a higher-tier model with a larger context window so all 14 files receive adequate attention in one pass.
• B.Require developers to manually split large PRs into smaller submissions of 3-4 files before triggering the automated review.
• C.Run three independent single-pass reviews on the full PR and only flag issues that achieve consensus across at least two runs.
• D.Split the review into focused passes: analyze each file individually for local issues, followed by a separate integration-focused pass for cross-file data flow.

Comment: *

Name: *

Email: *

Verification: *

Your get_portfolio_value tool returns the total value of a user's investment portfolio. You're deciding between returning a structured JSON object with explicit fields versus returning the information as a formatted text string. What is the primary advantage of using structured output with defined fields?

• A.JSON schemas automatically validate that the underlying API returned correct data before the agent processes it.
• B.Structured JSON consumes significantly fewer tokens than natural language, substantially reducing API costs.
• C.Structured JSON is processed deterministically by the model, significantly improving accuracy when extracting values.
• D.The agent can reliably extract specific values without parsing free-form text, reducing errors in subsequent operations.

Comment: *

Name: *

Email: *

Verification: *