Question 280
A team's policy script exits with a blocking status before a destructive command. How should the agent runtime respond?
- Produce a blocking control result instead of an informational log event
- Reject closed or suspend the protected action according to explicit policy
- Block over-budget requests in the tool-call authorization boundary
- Honor the hook's documented block decision and return the reason to the agent
Explanation
Correct Answer
D:A hook decision is part of the deterministic policy boundary for tool execution. Swallowing a deny signal would make the guard ineffective and hide the cause from the agent.
Incorrect Answers
A:Emitting telemetry alone does not enforce permission decisions; its benefits for distinguishing logging from access denial do not resolve the need for respecting a hook's deny outcome.
B:When the decision service is unavailable, a high-risk action should not silently proceed; supporting handling a failed mandatory policy check cannot replace the required respecting a hook's deny outcome.
C:Tracking usage and checking it before execution makes a budget enforceable; this contributes to enforcing a quota at the tool boundary without meeting the need for respecting a hook's deny outcome.
Question 281
A long conversation contains finished tool calls and pending follow-ups. What must a reliable checkpoint preserve?
- Focus on forking from the same validated baseline and keep branch histories separate
- Launch a separate session with its own customer and authorization context
- Confirm the persisted operation ID against the payment system before retrying
- Store message order and essential tool-result linkage along with session metadata
Explanation
Correct Answer
D:Resumable state includes structured execution history, not merely the final natural-language summary. Losing invocation linkage makes retries and interpretation unreliable.
Incorrect Answers
A:For comparing alternatives from a common session baseline, a common starting context makes alternatives comparable; however, this requirement concerns checkpointing structured agent progress.
B:Separate sessions prevent accidental disclosure and policy contamination between customers; that may serve isolating unrelated customer conversations, but it is no substitute for checkpointing structured agent progress.
C:A durable transcript does not guarantee whether the external effect committed; this may support reconciling external actions after session interruption but cannot alone meet the requirement for checkpointing structured agent progress.
Question 282
A security review is written by the same agent that produced the code and regularly misses its own assumptions. What architecture reduces bias?
- Fix validated facts and limit later passes to approved presentation edits
- Include independent evidence checks rather than relying solely on answer agreement
- Assemble a labeled test set including true positives, negatives and borderline cases
- Use an independent review pass with source requirements and a separate rubric
Explanation
Correct Answer
D:Independent evaluation can challenge design assumptions rather than merely reaffirm the author. Separation of role and context makes blind spots more visible.
Incorrect Answers
A:Separating factual content from stylistic changes reduces accidental meaning drift; the option would serve preserving verified facts during editing, not improving independence between author and reviewer as specified in the stem.
B:Several models may share the same misconception; this addresses avoiding false confidence from reviewer agreement, while the question is about improving independence between author and reviewer.
C:Calibration requires examples with known outcomes, not only successful demonstrations; that relates to measuring review precision against ground truth, not the necessary improving independence between author and reviewer.
Question 283
The team wants to investigate every tool error without granting subagents access to raw backend logs. What allocation is appropriate?
- Let a privileged diagnostics component inspect logs and return sanitized findings
- Present separated tool sets or filtered capabilities according to worker identity
- Supply the coordinator lookup access and confine refund mutations to the authorized specialist
- Delete the email-sending tool while keeping read and draft capabilities
Explanation
Correct Answer
A:Dedicated diagnostics isolates sensitive infrastructure data while providing actionable summaries. Granting logs access to all workers would expand the attack surface unnecessarily.
Incorrect Answers
B:Different trust levels deserve different callable surfaces; the unsatisfied requirement is segregating diagnostic access from regular agents, not partitioning sensitive and ordinary MCP capabilities.
C:Partitioning capabilities prevents the orchestration role from bypassing specialist approval; in this scenario, isolating privileged refund capability is secondary to segregating diagnostic access from regular agents.
D:When considering separating drafting from outbound delivery permissions, tool distribution should distinguish generating a proposed communication from transmitting it; the requirement for segregating diagnostic access from regular agents remains unmet.
Question 284
A task asks the agent to delete unused files, but no usage analysis has been performed. Which mode is safer?
- Investigate references and present a removal plan before deletion
- Begin in plan mode for investigation and architectural approval
- Declare deployment order, checkpoints and rollback triggers before execution
- Assemble performance evidence and define the target metric before code changes
Explanation
Correct Answer
A:Deleting files on a hunch can break hidden consumers. Discovery and approval establish whether the files are truly safe to remove.
Incorrect Answers
B:Plan mode helps explore the codebase and propose changes without immediately applying edits; this concerns planning before a cross-cutting implementation, leaving the requirement for researching destructive refactors before execution outstanding.
C:Multi-service updates have dependencies and failure domains that cannot be improvised safely; preparing release sequencing and rollback and researching destructive refactors before execution are distinct responsibilities under these constraints.
D:Optimization without a baseline makes success impossible to judge and may harm unrelated paths; the action supports baselining performance before optimization but falls short of researching destructive refactors before execution in this case.
Question 285
A form is unreadable because the scan contains a missing page. How should the pipeline respond?
- Request a better source or human review instead of repeated identical retries
- Use a bounded retry budget and move the document to an exception queue
- Refrain or mark the amount uncertain pending human verification
- Verify extracted references against authoritative business records
Explanation
Correct Answer
A:An absent source page cannot be recovered by sampling the same incomplete input. Escalation is more trustworthy than repeatedly inventing absent content.
Incorrect Answers
B:Repeated retries cost money and may not improve fidelity; that serves bounding extraction repair attempts, whereas the case requires recognizing a nonrecoverable extraction input gap.
C:Because high-impact numeric ambiguity deserves explicit uncertainty, the measure supports calibrating abstention for unreadable amounts rather than the needed recognizing a nonrecoverable extraction input gap.
D:Referential checks establish that identifiers and relationships are real; its benefits for checking referential integrity after extraction do not resolve the need for recognizing a nonrecoverable extraction input gap.
Question 286
A bug report gives an exact path and line range to inspect, but the agent starts recursive scanning. What is the efficient correction?
- Read the indicated path and relevant region directly
- Perform the formatter through Bash and inspect the resulting diff or file changes
- Adopt Grep to locate matching text across the repository
- Choose Write to create the required file content
Explanation
Correct Answer
A:A precise location is already supplied, so broad search adds noise and latency. Direct inspection better uses the available evidence and context budget.
Incorrect Answers
B:Process output alone may not prove exactly what changed; the stated problem requires avoiding unnecessary repository-wide discovery rather than verifying changes produced by an executable command.
C:Grep searches contents and identifies files or lines containing a pattern; a method suited to finding a phrase inside repository files does not meet the need for avoiding unnecessary repository-wide discovery.
D:For creating a new source-controlled document, write creates or replaces file contents at a chosen path; however, this requirement concerns avoiding unnecessary repository-wide discovery.
Question 287
An escalation request goes to a human queue without the transcript or reason code. What should the handoff include?
- Begin by making eligibility verification a required predecessor in the workflow graph
- Serialize or lock the conflicting state-changing section
- Attach a concise case summary, relevant evidence and the escalation reason
- Express external-input and tool-pending states separately
Explanation
Correct Answer
C:A meaningful handoff preserves the context the reviewer needs to act, avoiding repeated questioning and incorrect policy decisions.
Incorrect Answers
A:An explicit dependency ensures the mutation cannot begin before policy checks succeed; that relates to enforcing refund prerequisites before mutation, not the necessary transferring sufficient escalation context.
B:Although concurrency control protects shared mutable state, the choice addresses preventing conflicting concurrent mutations and not the requested transferring sufficient escalation context.
D:Different waits have different timeouts, retries and authorization requirements; the option would serve distinguishing human and tool handoff states, not transferring sufficient escalation context as specified in the stem.
Question 288
An agent frequently chooses the wrong search tool because two descriptions promise to retrieve customer information. What helps most?
- Enforce a scoped tool allowlist for the worker invocation
- Make available separated tool sets or filtered capabilities according to worker identity
- Clarify each tool's selection criteria and record type in descriptions
- Offer the coordinator lookup access and confine refund mutations to the authorized specialist
Explanation
Correct Answer
C:Tool routing quality depends on distinct metadata. Differentiating CRM profiles from order histories lets the model choose the right evidence source without guessing.
Incorrect Answers
A:A narrow allowance reduces both accidental mutation and prompt-injection opportunities; in this scenario, applying per-worker tool allowlists is secondary to disambiguating adjacent retrieval tools.
B:When considering partitioning sensitive and ordinary MCP capabilities, different trust levels deserve different callable surfaces; the requirement for disambiguating adjacent retrieval tools remains unmet.
D:Partitioning capabilities prevents the orchestration role from bypassing specialist approval; isolating privileged refund capability and disambiguating adjacent retrieval tools require different controls in this situation.
Question 289
A model formats extracted dates inconsistently across regions. What example design supports normalization?
- Present labeled cases where multiple classes are present and document the priority rule
- Start by making examples conform to a single documented schema and labeling convention
- Show diverse input date formats paired with one required normalized representation
- Embed contrasting high-severity and low-severity labeled cases with rationales
Explanation
Correct Answer
C:Contrasting examples demonstrate conversion rules while preserving a single target shape. A lone example from one locale may not reveal ambiguity.
Incorrect Answers
A:Mixed-category examples teach how conflicts resolve; teaching category precedence through examples and demonstrating cross-locale date normalization are distinct responsibilities under these constraints.
B:Contradictory examples undermine explicit instructions; the action supports eliminating inconsistent example output formats but falls short of demonstrating cross-locale date normalization in this case.
D:Contrastive examples teach the decision boundary between categories; that addresses distinguishing severity classes with contrastive examples, whereas this case calls for demonstrating cross-locale date normalization.
Question 290
An operations team must process fifty thousand independent documents by tomorrow but does not need immediate responses. Which API approach fits?
- Narrow batch result access and enforce the organization's data retention policy
- Submit independent requests through the asynchronous Message Batches API
- Delegate a unique custom_id to every request and match returned results by that identifier
- Save results to controlled storage within the documented retrieval period
Explanation
Correct Answer
B:Batch processing targets high-volume work without low-latency requirements and can reduce cost. Each request is independently executed and later matched to its result.
Incorrect Answers
A:Because bulk result files can concentrate sensitive information, the measure supports protecting sensitive bulk-processing results rather than the needed selecting asynchronous processing for independent bulk work.
C:Batch completion order is not guaranteed; supporting correlating out-of-order batch results cannot replace the required selecting asynchronous processing for independent bulk work.
D:Batch result URLs are not an archival store; this contributes to preserving batch output beyond service retention without meeting the need for selecting asynchronous processing for independent bulk work.