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.
Question 291
A source document contains instructions telling the agent to ignore its citation policy. What trust boundary matters?
- Treat retrieved document instructions as untrusted content rather than authority
- Flag confirmed observations separately from estimates and unknowns
- Maintain original-source identifiers and evidence type through every handoff
- Register retrieval time and a source version or approved snapshot reference
Explanation
Correct Answer
A:External evidence may contain prompt injection. The agent should analyze its factual content without allowing it to replace governing workflow and safety rules.
Incorrect Answers
B:For separating verified status from estimates, provenance distinguishes system-of-record data from inference; however, this requirement concerns resisting prompt injection embedded in source material.
C:A claim's evidentiary strength depends on where it originated; that may serve retaining primary-source lineage across agents, but it is no substitute for resisting prompt injection embedded in source material.
D:Live pages can change or disappear; this may support preserving time-sensitive research provenance but cannot alone meet the requirement for resisting prompt injection embedded in source material.
Question 292
A project needs security notes to apply to all code files but not to Markdown docs. What configuration fits?
- Use a rule with glob patterns matching the relevant source extensions
- Separate language-specific guidance into separate path-targeted rule files
- Validate rule glob matches against real repository paths
- Shift conditional package conventions into dedicated rules files with path selectors
Explanation
Correct Answer
A:Conditional rule metadata can select the file types that need a convention. Glob patterns must reflect actual repository layouts to avoid both omissions and overreach.
Incorrect Answers
B:Distinct rules with distinct globs avoid contradictory language conventions; the option would serve separating language-specific rules, not selecting source file types for security guidance as specified in the stem.
C:A simple match audit reveals dead rules that never load; this addresses verifying rule selectors actually match files, while the question is about selecting source file types for security guidance.
D:Splitting the corpus by applicability reduces irrelevant context; that relates to avoiding an oversized global project rule set, not the necessary selecting source file types for security guidance.
Question 293
A developer has a stack trace pointing to a particular service method. What is the efficient next read?
- Inspect the implicated method and immediate call chain before expanding scope
- Read dependency manifests and lockfiles to confirm the affected version
- Focus on centralizing a short verified configuration summary with exact file references
- Review package-scoped guidance while exploring the affected subtree
Explanation
Correct Answer
A:Error evidence supplies a high-signal starting point. Focused inspection avoids spending tokens on unrelated modules unless the first hypothesis fails.
Incorrect Answers
B:Current local package metadata grounds the remediation plan; grounding dependency remediation in repository metadata and using stack traces to prioritize code inspection require different controls in this situation.
C:Shared grounded notes prevent repeated expensive reads while retaining the ability to verify details; the unsatisfied requirement is using stack traces to prioritize code inspection, not reusing concise verified repository findings.
D:Local rules capture framework and testing differences across packages; in this scenario, using local conventions during monorepo exploration is secondary to using stack traces to prioritize code inspection.
Question 294
A tool must reject an amount value given as a string instead of a number. What contract should be declared?
- Pair JSON-schema checking with validation against the source facts
- Define line items as an array with a defined item object schema
- Adopt nested object definitions with typed address properties
- Define the tool input schema with a numeric amount field and strict validation where supported
Explanation
Correct Answer
D:Typed schema constraints differentiate numeric amounts from text. Strict tool-use support can enforce input shape while business checks still validate the actual transaction.
Incorrect Answers
A:A schema constrains representation, not factual truth; this can serve distinguishing schema conformance from factual accuracy, but leaves the requirement for enforcing typed arguments for a tool call unmet.
B:The declared array type tells the consumer how to iterate through repeated records; that addresses constraining collection shape in extraction output, whereas this case calls for enforcing typed arguments for a tool call.
C:A nested schema preserves structural relationships and makes each field independently verifiable; this concerns constraining nested JSON extraction structures, leaving the requirement for enforcing typed arguments for a tool call outstanding.
Question 295
A support agent should approve routine refunds automatically when within policy limits, but route over-limit cases for authorization. Which workflow design enforces that conditional boundary?
- Add an executable policy gate before the final approval or refund call
- Build in a test-pass acceptance condition before marking the coding task done
- Document a post-action reconciliation stage before marking the case resolved
- Send every refund request to a human before processing regardless of the amount or policy threshold
Explanation
Correct Answer
A:A deterministic guard can compare amounts with configured limits. Relying only on natural-language reminders risks occasional unauthorized action.
Incorrect Answers
B:Stage completion is a business invariant that should be checked with tool output; this contributes to using tests as a completion gate without meeting the need for enforcing financial limits in a workflow.
C:Workflow success should depend on confirmed downstream state, not just a tool invocation; that serves verifying all required stages of case resolution, whereas the case requires enforcing financial limits in a workflow.
D:Mandatory review of all requests can stop unauthorized refunds but defeats the requirement to apply a threshold-sensitive route. A policy gate can allow eligible routine refunds and escalate only over-limit cases.
Question 296
A customer support assistant detects a credible account takeover risk during an ordinary order inquiry. What routing rule fits?
- Log the exception rationale and route it through the authorized review process
- Deliver a concise structured handoff with issue, evidence, attempted steps and decision needed
- Escalate to the security or fraud workflow and suspend sensitive actions
- Declare the approval dependency and avoid claiming the refund has occurred
Explanation
Correct Answer
C:Fraud indicators change the permitted handling of customer accounts. Continuing routine self-service could worsen harm or leak protected information.
Incorrect Answers
A:Unusual circumstances may justify discretion but need a human or policy-sanctioned decision; that may serve processing policy exceptions through approval, but it is no substitute for routing suspected fraud out of standard support.
B:Structured escalations let the human act on the actual unresolved choice; this may support preparing an actionable human escalation but cannot alone meet the requirement for routing suspected fraud out of standard support.
D:Transparent status respects the user's need for clarity without misrepresenting incomplete action; a method suited to communicating a pending human approval accurately does not meet the need for routing suspected fraud out of standard support.
Question 297
OCR text omits the invoice number, but the model supplies a plausible identifier. What should the pipeline do?
- Validate extracted items against their source spans or document evidence
- Confirm extracted references against authoritative business records
- Re-execute with targeted feedback identifying the invalid field and expected format
- Maintain the original string and flag ambiguity until locale evidence resolves it
- Mark the missing field as unresolved rather than invent a value
Explanation
Correct Answer
E:Extraction is grounded in source content. A plausible format is not evidence that a specific number appeared in the document.
Incorrect Answers
A:Per-field provenance validation detects additions not supported by the original content; this addresses matching extracted entities to source evidence, while the question is about preventing fabricated missing source values.
B:Referential checks establish that identifiers and relationships are real; that relates to checking referential integrity after extraction, not the necessary preventing fabricated missing source values.
C:Although feedback-based repair gives the model a specific correctable defect, the choice addresses targeting a retry at a correctable parse defect and not the requested preventing fabricated missing source values.
D:A date parser cannot safely invent locale rules; the deciding condition is preventing fabricated missing source values, whereas this action concerns handling ambiguous date interpretations.
Question 298
A coordinator must avoid letting a search-only worker submit customer refunds. What permission design fits?
- Restrict the worker's allowedTools to the read-only search surface
- Request evidence of which documents and versions the worker actually consulted
- Start the two workers in separate contexts and compare at the parent level
- Constrain both delegated instructions and available data connectors to the assigned case
Explanation
Correct Answer
A:A narrow delegated tool allowlist prevents accidental or malicious state changes. Merely labeling a worker 'searcher' does not enforce its capabilities.
Incorrect Answers
B:Source and version metadata distinguish inspected facts from memory-based guesses; in this scenario, validating the evidentiary basis of a subagent answer is secondary to granting least-privilege tools to a worker.
C:When considering obtaining independent subagent hypotheses, isolated contexts create genuinely independent assessments; the requirement for granting least-privilege tools to a worker remains unmet.
D:When addressing narrowing a worker's data-access boundary, context minimization and access controls complement each other; the present task instead calls for granting least-privilege tools to a worker.
Question 299
A cross-service change plan lacks both test-based acceptance criteria and release sequencing with rollback checkpoints. Which TWO additions make the plan operational? Choose TWO.
- Document the release sequence with explicit checkpoints and rollback triggers
- Tie planned cross-service edits to testable acceptance criteria and named verification steps
- Compile performance evidence and define the target metric before code changes
- Map references and present a removal plan before deletion
Explanation
Correct Answers
A:Alongside embedding verification into the implementation plan, the workflow needs preparing release sequencing and rollback: multi-service updates have dependencies and failure domains that cannot be improvised safely; A planned sequence gives operators explicit hold and recovery points.
B:A useful plan describes not just changes but the evidence that will establish success. Without verification steps, implementation can terminate after plausible-looking edits.
Incorrect Answers
C: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 embedding verification into the implementation plan in this case.
D:Deleting files on a hunch can break hidden consumers; this can serve researching destructive refactors before execution, but leaves the requirement for embedding verification into the implementation plan unmet.
Question 300
A policy-sensitive contract extractor fails on rare jurisdictions and faces unclear exceptions requiring an authorized decision. Which TWO review actions are appropriate? Choose TWO.
- Refer undocumented jurisdictional exceptions to an authorized policy owner before commitment
- Sample unfamiliar, high-risk jurisdictions for deeper human review rather than auto-acceptance
- Connect approval to immutable action parameters and require reapproval after material changes
- Auto-approve unfamiliar jurisdictions when they appear infrequently, since a small cohort cannot affect overall accuracy much
Explanation
Correct Answers
A:Routing uncertain policy exceptions to authorized review matters in this scenario: policy ambiguity is not resolved by fluent reasoning; A controlled exception channel lets a human decide without making an unauthorized promise.
B:Aggregate accuracy can hide weak performance on small important cohorts. Risk-aware stratified checks expose failures that a uniform random sample might miss.
Incorrect Answers
C:Approval of one transaction should not authorize a modified one; its benefits for tying human approval to exact mutation details do not resolve the need for targeting review to difficult high-impact document cohorts.
D:Low-volume jurisdictions can carry disproportionate regulatory risk. Novel or unclear exceptions should be escalated rather than accepted on global averages.