Question 210 A tool calls an external service that is throttling. What should a structured response preserve? A fake successful record to hide capacity problems. Only a generic “bad” string. An instruction to remove tool request limits globally. A recognizable rate-limit category and a safe retry-after interval when available. Explanation Correct Answer D:The caller can delay and cap future calls rather than hammering the backend. Incorrect Answers A:False output pollutes downstream results. B:The agent cannot distinguish waiting from correcting its input. C:Bypassing rate limits is not a valid recovery.
Question 211 A coordinator is designing a safety assessment with one data-collection worker, one hazard analyst, and one report writer. Which TWO responsibilities belong with the coordinator? Have every worker alter all other workers’ instructions at will. Check coverage and conflicts before releasing the integrated report. Remove provenance before synthesis to reduce context size. Define the task dependencies and the scope each specialist must cover. Delegate final approval authority to the fastest worker automatically. Explanation Correct Answers B:Integration quality requires reconciliation across worker outputs. D:Only the coordinator sees the end-to-end objective and can assign work without gaps. Incorrect Answers A:Uncoordinated authority changes make the workflow hard to govern. C:A report cannot defend its claims if evidence references are lost. E:Completion speed is not a substitute for system-wide validation.
Question 212 A team wants a reviewer to focus on authorization flaws rather than formatting. What is the most relevant prompt boundary? Allow the model to report unrelated style changes as security findings. Ask the model to count whitespace characters first. Define authorization invariants and require a concrete bypass trace before flagging. Increase severity for every missing comment. Explanation Correct Answer C:Permission boundaries create a measurable security target. Incorrect Answers A:Mixing issue types reduces signal quality. B:Whitespace rarely proves an authorization bypass. D:Documentation gaps are not automatically access-control vulnerabilities.
Question 213 A team wants Claude Code to generate test suggestions in CI but not commit changes. Which boundary supports this? Grant full push access because test suggestions may be useful. Have the model directly modify production secrets. Run in read-only mode and publish structured suggestions as artifacts or comments. Disable all version control audit records. Explanation Correct Answer C:The task is analysis, not mutation of the branch. Incorrect Answers A:Suggestions do not require write permission. B:That is unrelated and unsafe. D:Visibility is important in CI.
Question 214 A tool schema permits arbitrary file paths, but its purpose is to read reports in one designated folder. What should an architect do? Hide the path argument in an unstructured text field. Trust the model never to request out-of-folder paths. Constrain permitted paths server-side and document the accepted scope. Return other customers’ files if the path looks plausible. Explanation Correct Answer C:Tool descriptions help selection, but real path authorization must be enforced by the service. Incorrect Answers A:Unstructured input makes validation more difficult. B:Prompt-following cannot replace an access control check. D:Unscoped paths create cross-tenant data exposure.
Question 215 A workflow produces a generated press release before legal review completes. What should control the handoff to publication? A random short delay after drafting. The presence of the word “approved” anywhere in the draft. The number of paragraphs in the press release. A machine-checked approval state from the legal review step. Explanation Correct Answer D:Publication depends on completed review, not generated prose asserting approval. Incorrect Answers A:Time elapsed cannot prove the required reviewer acted. B:The model could quote that word without a real review. C:Document length is unrelated to legal authorization.
Question 216 A second model approves a generated answer because the first model supplied its preferred answer key. How should that review change? Tell the checker that the supplied key is guaranteed correct. Judge the stem and options independently before revealing the proposed key, then compare rationales. Ignore option ambiguity and verify spelling only. Accept the key whenever the drafting agent gives a detailed explanation. Explanation Correct Answer B:Blinded review reduces anchoring toward an incorrect original key. Incorrect Answers A:That defeats the purpose of verification. C:Answer exclusivity is central to multiple-choice quality. D:Plausible prose can defend a wrong answer.
Question 217 A Claude Code agent is asked to locate the authorization check in a very large monorepo. What is the most efficient first pass? Assume the authorization logic lives in the first file named auth. Read every source file in alphabetical order before searching. Rewrite all middleware based on an initial guess. Use targeted filename and symbol searches to map candidate modules, then read relevant code paths. Explanation Correct Answer D:Narrow discovery preserves context while locating the actual enforcement point. Incorrect Answers A:Names are hints, not proof of runtime flow. B:Full-corpus reading exhausts context and delays the answer. C:A location investigation does not authorize mass edits.
Question 218 A user wants every file under docs/api/ that ends in .md. Which operation best scopes discovery? Edit README.md with a guessed list. Bash to rewrite directory permissions. Glob with the docs/api/**/*.md pattern. Grep for the string .md in source text. Explanation Correct Answer C:Path matching returns the relevant markdown files efficiently. Incorrect Answers A:Guessing file inventory does not discover actual paths. B:Permissions do not answer the path query. D:Content search can miss markdown filenames entirely.
Question 219 A coordinator asks two specialists for competing cost estimates. Which TWO handoff fields enable a valid comparison? The specialist’s nickname as the sole identifier. A merged total with no source distinction. An instruction to omit confidence and limitations. The assumptions and unit basis behind each estimate. A link to the underlying calculations or source rows. Explanation Correct Answers D:Costs cannot be compared if one includes labor and the other excludes it. E:Traceable calculations permit reconciliation of conflicting figures. Incorrect Answers A:A nickname does not state what was measured. B:Premature merging prevents the coordinator from auditing each estimate. C:Unknown inputs and risks matter to cost interpretation.
Question 220 A batch pipeline is overwhelmed because 50,000 documents each contain long repeated policies. Which pre-processing design is appropriate? Deduplicate stable reference material and supply only the relevant context for each item. Add the full document collection to every request. Remove all source context and force the model to guess. Increase concurrent writes without regard to API limits. Explanation Correct Answer A:Reducing redundant input saves tokens while preserving necessary evidence. Incorrect Answers B:That increases cost and contamination. C:That harms grounded accuracy. D:Unbounded concurrency may produce more failures.
