Question 50
A system requires structured extraction but sometimes receives a model refusal. What should its parser account for?
- Choose nested object definitions with typed address properties
- Reduce the schema to supported constructs and validate remaining rules in application code
- Track versions of the result contract and test consumers before deploying schema changes
- Detect nonstandard stop or refusal outcomes before treating content as a schema result
Explanation
Correct Answer
D:Schema-oriented APIs may have special refusal or termination paths. Robust consumers inspect response status instead of assuming every completion is a usable record.
Incorrect Answers
A:A nested schema preserves structural relationships and makes each field independently verifiable; that serves constraining nested JSON extraction structures, whereas the case requires handling non-result terminations in structured workflows.
B:Because constrained decoding has a documented supported JSON Schema subset, the measure supports respecting structured-output schema limitations rather than the needed handling non-result terminations in structured workflows.
C:Structured outputs create a stable interface that downstream systems depend upon; its benefits for managing downstream compatibility of structured outputs do not resolve the need for handling non-result terminations in structured workflows.
Question 51
A large extraction run would benefit from lower processing cost but can tolerate delayed completion. What performance trade-off is appropriate?
- Allocate a unique custom_id to every request and match returned results by that identifier
- Prefer asynchronous batching and budget for deferred results
- Track pending request limits and pace submissions by available capacity
- Employ a synchronous interactive request for the latency-sensitive turn
Explanation
Correct Answer
B:Batch processing can reduce unit cost at the expense of immediate completion. It is a deliberate throughput-versus-latency choice rather than a universal replacement for online requests.
Incorrect Answers
A:Batch completion order is not guaranteed; the stated problem requires trading response latency for bulk-processing cost rather than correlating out-of-order batch results.
C:For respecting throughput limits when batching, batch queues remain subject to provider limits and account constraints; however, this requirement concerns trading response latency for bulk-processing cost.
D:A batch job runs asynchronously and completion is not immediate; that may serve keeping interactive requests out of deferred batches, but it is no substitute for trading response latency for bulk-processing cost.
Question 52
Three reviewers see the same generated draft and agree by majority vote even though the draft may contain fabricated statements. What validation weakness remains?
- Add independent evidence checks rather than relying solely on answer agreement
- Ask the reviewer to propose evidence-backed findings and leave changes to an authorized implementer
- Lock validated facts and limit later passes to approved presentation edits
- Give all reviewers the same generated draft and ask them whether it seems credible
- Accept the majority opinion as factual proof without linking the claims to inspected primary records
Explanation
Correct Answer
A:Several models may share the same misconception. Corroboration from tests and source documents is stronger than consensus without external evidence.
Incorrect Answers
B:Reviewer and implementer responsibilities differ; the deciding condition is avoiding false confidence from reviewer agreement, whereas this action concerns separating review conclusions from edit authority.
C:Separating factual content from stylistic changes reduces accidental meaning drift; the option would serve preserving verified facts during editing, not avoiding false confidence from reviewer agreement as specified in the stem.
D:Shared unverified text creates correlated errors: all reviewers can echo the same invented claim. Independent source verification is the needed check, not confidence in the common draft.
E:A vote measures agreement, not truth; correlated hallucinations can win unanimously. Source-grounded validation has to be independent of answer agreement.
Question 53
A team wants to format generated source files immediately after edits rather than only at task end. Which lifecycle event fits?
- Restrict the hook matcher to the intended tool name or invocation pattern
- Inspect the tool completion event and record its returned outcome
- Run a PostToolUse handler after successful file-edit tools
- Check and safely handle hook inputs before invoking external processes
Explanation
Correct Answer
C:A post-tool callback can inspect the completed edit and trigger formatting or verification in response. It acts after the operation instead of preemptively approving it.
Incorrect Answers
A:When considering targeting a hook to relevant operations, matchers determine which hook executions apply; the requirement for reacting to a completed editing tool remains unmet.
B:When addressing auditing tool execution outcomes, post-execution telemetry can contain the result and status of a tool operation; the present task instead calls for reacting to a completed editing tool.
D:Hooks execute application-controlled code and may receive adversarial strings; the unsatisfied requirement is reacting to a completed editing tool, not validating data inside enforcement hooks.
Question 54
A team wants new contributors to know which scripts validate code before PR submission. What belongs in project guidance?
- Document the repository's actual build, lint and test commands with prerequisites
- Check the intended CLAUDE.md is discovered and loaded for that working context
- Eliminate secrets from instruction text and use secure environment-based credential provisioning
- Place reviewed repository conventions in the project-level CLAUDE.md
Explanation
Correct Answer
A:Executable command knowledge is stable operational context. Recording real prerequisites helps the agent validate changes without guessing project tooling.
Incorrect Answers
B:Rules are useful only when the assistant encounters the relevant file and scope; this can serve verifying instruction-file loading, but leaves the requirement for recording project-specific validation commands unmet.
C:Instruction files are often shared or exposed as model context; that addresses protecting credentials from instruction-file exposure, whereas this case calls for recording project-specific validation commands.
D:Project instruction files carry shared guidance alongside the codebase; this concerns distributing shared repository coding instructions, leaving the requirement for recording project-specific validation commands outstanding.
Question 55
An agent wants to determine whether a command-line formatter altered files. Which tool combination gives authoritative evidence?
- Use Grep to search occurrences, inspect affected files, then apply controlled Edits
- Choose Edit for a targeted modification to an existing file
- Choose Read to examine the specified file
- Run the formatter through Bash and inspect the resulting diff or file changes
Explanation
Correct Answer
D:Process output alone may not prove exactly what changed. Execution plus repository diff inspection reveals side effects and allows review before acceptance.
Incorrect Answers
A:A search-first workflow reveals scope and avoids blind global replacement; its benefits for scoping a multi-file refactor before mutation do not resolve the need for verifying changes produced by an executable command.
B:Targeted editing changes selected text while maintaining the surrounding source; supporting changing a limited region of existing code cannot replace the required verifying changes produced by an executable command.
C:Read provides file content for reasoning without modifying the repository; this contributes to inspecting a known file without edits without meeting the need for verifying changes produced by an executable command.
Question 56
Two tool definitions share the name get_status but refer to different business systems. What design change reduces routing ambiguity?
- Decline the unauthorized call and report a permission denial to orchestration
- Focus on setting per-task call budgets and terminate or escalate on exhaustion
- Namespace or rename tools to identify their owning integration and purpose
- Introduce a pre-execution authorization and query-validation layer at the tool dispatcher
Explanation
Correct Answer
C:Unique, descriptive identifiers help the agent and dispatcher target the intended backend. Bare generic names collide as the available tool catalogue grows.
Incorrect Answers
A:For enforcing a denied capability at runtime, a capability boundary must be enforced at execution time; however, this requirement concerns resolving tool name collisions.
B:Tool budgets prevent unbounded external work and let the system deliver a controlled incomplete status; that may serve controlling tool invocation cost per task, but it is no substitute for resolving tool name collisions.
D:Model-generated arguments are proposals; the stated problem requires resolving tool name collisions rather than validating privileged tool calls before execution.
Question 57
A high-value financial transfer is proposed by an agent. Which review design supplies real authorization?
- Contrast escalation outcomes with labeled ground truth and false-positive rates
- Require an authenticated human approval before the mutation is dispatched
- Retain individual rationales, evidence and the final adjudication basis
- Select samples by document type and route high-risk unfamiliar cases for deeper inspection
Explanation
Correct Answer
B:Approval must be checked at the controller boundary rather than inferred from model confidence. High-impact financial changes need explicit authorized consent.
Incorrect Answers
A:Thresholds should be calibrated empirically, not chosen solely from model self-descriptions; the option would serve calibrating escalation thresholds against observed outcomes, not enforcing human authorization for high-value actions as specified in the stem.
C:Disagreement may expose ambiguous policies or missing facts; that relates to retaining review disagreement and adjudication provenance, not the necessary enforcing human authorization for high-value actions.
D:Although aggregate accuracy can hide weak performance on small important cohorts, the choice addresses targeting review to difficult high-impact document cohorts and not the requested enforcing human authorization for high-value actions.
Question 58
A skill's documentation still describes a retired internal API. What release practice prevents broken reuse?
- Choose project-relative paths and document required environment inputs
- Organize the release workflow as a project skill with clear instructions
- Provide the skill a clear trigger description and intended use cases
- Version and test the skill instructions when the dependent API changes
Explanation
Correct Answer
D:Reusable automation is subject to dependency drift. Treating skills like maintained code prevents stale tool assumptions from recurring in every invocation.
Incorrect Answers
A:Portable skills avoid machine-specific assumptions and resolve resources from a predictable project context; removing machine-specific assumptions from a skill and maintaining version compatibility of a reusable skill require different controls in this situation.
B:A reusable skill can collect steps, supporting files and tool permissions under one invocable workflow; the unsatisfied requirement is maintaining version compatibility of a reusable skill, not making a repeatable release workflow reusable.
C:Descriptive metadata improves discovery and correct invocation; in this scenario, improving skill discoverability is secondary to maintaining version compatibility of a reusable skill.
Question 59
An agentic research system produces duplicated web searches across several workers. What coordinator improvement addresses the waste?
- Deduplicate planned search tasks and share a normalized evidence index
- Send cross-specialty requests through an orchestrator with an explicit task contract
- Introduce a deterministic aggregation policy for worker output acceptance
- Separate independent expert analyses and reserve integration for the orchestrator
Explanation
Correct Answer
A:Centralizing a work plan and results catalogue prevents workers from independently repeating identical queries. It also provides stable references for later synthesis.
Incorrect Answers
B:Hub-and-spoke mediation controls ownership, context, and failure propagation; this concerns minimizing unpredictable peer-to-peer dependencies, leaving the requirement for avoiding redundant delegated searches outstanding.
C:A merge contract can specify completeness, conflict handling and evidence precedence; defining a reproducible worker merge policy and avoiding redundant delegated searches are distinct responsibilities under these constraints.
D:Focused workers can receive targeted instructions and smaller context windows; the action supports matching specialist agents to distinct competencies but falls short of avoiding redundant delegated searches in this case.
Question 60
One worker needs another's normalized data before analysis begins. How should dependency planning represent this?
- Repeat or replace only the failed subtask while keeping successful outputs
- Make data normalization a prerequisite for the dependent analysis
- Specify measurable deliverables and pass-fail conditions for each subtask
- Partition the issue into bounded states with explicit exit and escalation criteria
Explanation
Correct Answer
B:The task graph must express actual data dependencies. Running both phases in parallel can force the analyst to guess a schema or use stale inputs.
Incorrect Answers
A:Independence in the task graph permits selective replay; this contributes to isolating retry to failed work without meeting the need for reflecting real data dependencies in the task graph.
C:Because a task boundary is useful when the parent can decide whether its output satisfies requirements, the measure supports specifying task-level acceptance criteria rather than the needed reflecting real data dependencies in the task graph.
D:Bounded tasks define what evidence and action terminate work; its benefits for defining bounded task exits do not resolve the need for reflecting real data dependencies in the task graph.