Question 550
Several clients cannot upgrade together when a tool input field is renamed. Which measure directly addresses the problem?
- Review post-incident root causes
- Remove deprecated configuration under change control
- Use automated integration tests before release
- Use backward-compatible interface evolution
Explanation
Correct Answer
D:Additive changes and versioned contracts allow old and new clients to coexist during migration; in production, an abrupt breaking rename can strand clients even if the model prompt is unchanged.
Incorrect Answers
A:Root-cause analysis traces initiating conditions and systemic prevention beyond short-term restoration; this addresses a separate concern, not the fact that additive changes and versioned contracts allow old and new clients to coexist during migration; a successful restart does not explain why the same fault reappears.
B:A controlled cleanup traces dependencies, verifies current usage, and deletes unused flags with tests; here, the controlling capability is that additive changes and versioned contracts allow old and new clients to coexist during migration, but removing all old switches at once can unintentionally disable required paths.
C:Integration tests check how independently coded components actually interact under realistic conditions; the missing capability is more specific: additive changes and versioned contracts allow old and new clients to coexist during migration; unit tests of isolated helpers cannot reveal all cross-component contract failures.
Question 551
The inference backend is down and front ends need a predictable response contract. How should the implementation handle this need?
- Send a typed fallback error to clients
- Use a circuit breaker for an unstable dependency
- Cap the number of concurrent requests
- Capture sanitized HTTP diagnostics
Explanation
Correct Answer
A:A well-defined error object lets clients distinguish outage, rate limit, and invalid input, subject to this requirement: returning fabricated success data makes incident diagnosis harder.
Incorrect Answers
B:Temporarily opening the breaker limits cascading failures and permits controlled recovery testing; here, the controlling capability is that a well-defined error object lets clients distinguish outage, rate limit, and invalid input, but it cannot fix a bug in the caller's own payload format.
C:A concurrency pool smooths demand and limits rate-limit errors; the missing capability is more specific: a well-defined error object lets clients distinguish outage, rate limit, and invalid input; throughput capacity and latency vary by model and account limits.
D:Structured redacted logs reveal failure context without leaking authorization headers; by comparison, the intended solution relies on this: a well-defined error object lets clients distinguish outage, rate limit, and invalid input; recording entire raw requests can expose customer secrets.
Question 552
An operations team must correlate a reported failure with the model, request path, and completion state. What is the right implementation response?
- Restart with a concise summary when context is exhausted
- Read usage counts from the response
- Inspect stop_reason before returning output
- Store response metadata with the trace
Explanation
Correct Answer
D:Structured metadata and correlation records provide context for debugging and cost attribution—with this limitation: logging only the visible answer removes information needed to identify stop or usage anomalies.
Incorrect Answers
A:Compaction retains goals, commitments, and key evidence in less context before the next call; the needed capability is different—structured metadata and correlation records provide context for debugging and cost attribution—and summarization can lose details, so critical structured state remains in application storage.
B:Response usage fields provide the token measures used for billing and diagnostics, whereas the requirement calls for another behavior: structured metadata and correlation records provide context for debugging and cost attribution; character count is not an exact substitute for model tokenization.
C:The completion reason tells the client whether to deliver, dispatch, or handle truncation; however, the critical mechanism here is that structured metadata and correlation records provide context for debugging and cost attribution, and treating all returned content as a complete final answer can hide partial results.
Question 553
An agent makes multiple reasoning calls and tool follow-ups to resolve one customer issue. What should be implemented to solve this problem?
- Track separate model pricing assumptions
- Aggregate usage across multi-step tasks
- Alert on unusual token spikes
- Compute cost from input and output token usage
Explanation
Correct Answer
B:Summing usage across the whole trace shows the true cost of satisfying one business request—with this limitation: only counting the final call omits earlier orchestration expenses.
Incorrect Answers
A:Per-tier pricing and request distribution are needed for realistic cost forecasts; however, the critical mechanism here is that summing usage across the whole trace shows the true cost of satisfying one business request, and a single blended rate can conceal costly high-tier escalation patterns.
C:Cost monitoring can detect and halt anomalous demand before it causes significant overspend; this cannot substitute for the behavior that summing usage across the whole trace shows the true cost of satisfying one business request, because monthly billing review alone detects runaway costs too late.
D:Per-model input and output token rates combine with measured usage for accurate costing; in contrast, summing usage across the whole trace shows the true cost of satisfying one business request, while number of messages alone is not an adequate unit of resource consumption.
Question 554
A shipment lookup cannot succeed without a warehouse ID and a shipment reference. Which option best fits the described workflow?
- Reconcile all returned tool results before synthesis
- Represent allowed values as enums where appropriate
- Constrain nested object shapes
- Mark truly required arguments in the schema
Explanation
Correct Answer
D:The schema identifies mandatory fields so calls missing them can be rejected before dispatch, subject to this requirement: optional property defaults should not hide missing business-critical identifiers.
Incorrect Answers
A:The harness maps all available results to their calls before requesting a comparative answer; the requirement instead turns on this behavior: the schema identifies mandatory fields so calls missing them can be rejected before dispatch; ignoring a pending result can produce an unsupported recommendation.
B:Constrained value sets prevent unsupported strings from reaching backend business logic; this cannot substitute for the behavior that the schema identifies mandatory fields so calls missing them can be rejected before dispatch, because valid enum membership alone cannot authorize an illegal state transition.
C:Structured nested validation rejects incomplete or malformed records consistently; in contrast, the schema identifies mandatory fields so calls missing them can be rejected before dispatch, while accepting arbitrary object shapes can produce unpredictable writes.
Question 555
A model output parser works syntactically but generated labels are often wrong. How should this requirement be met?
- Use a sandbox backend for tool integration tests
- Use a compatibility test matrix for upgrades
- Run contract tests against a versioned API schema
- Evaluate semantic behavior with labeled fixtures
Explanation
Correct Answer
D:A test set with expected outcomes measures content quality, not only code execution; its operational boundary is that a green HTTP response cannot establish answer correctness.
Incorrect Answers
A:A nonproduction environment exercises authorization, payload mapping, and failures safely; however, the critical mechanism here is that a test set with expected outcomes measures content quality, not only code execution, and faking every backend call can hide actual integration incompatibilities.
B:Matrix tests exercise supported combinations against contract and quality checks; the requirement instead turns on this behavior: a test set with expected outcomes measures content quality, not only code execution; assuming backward compatibility without testing risks hidden behavioral changes.
C:Contract tests enforce agreed field names, types, and error shapes across producer and client; this cannot substitute for the behavior that a test set with expected outcomes measures content quality, not only code execution, because casual manual testing may miss a breaking field change.
Question 556
Several clients exceed the service's permitted throughput and requests begin receiving rate-limit errors. Which measure directly addresses the problem?
- Reproduce with a raw valid API request
- Treat 400-class validation responses as request defects
- Treat HTTP 429 as throttling pressure
- Treat 5xx as potential upstream service faults
Explanation
Correct Answer
C:The caller reduces concurrency and uses bounded backoff while considering available quota; the important constraint is that blind parallel retries amplify rather than relieve throttle pressure.
Incorrect Answers
A:An isolated direct call distinguishes transport and wrapper behavior from a more complex integration path; the needed capability is different—the caller reduces concurrency and uses bounded backoff while considering available quota—and a successful API response does not automatically validate downstream application logic.
B:The developer checks the request schema, supported fields, and types before attempting another call, whereas the requirement calls for another behavior: the caller reduces concurrency and uses bounded backoff while considering available quota; retrying an identical malformed request wastes time and capacity.
D:The client applies guarded retries and monitoring while preserving safe degraded behavior; the requirement instead turns on this behavior: the caller reduces concurrency and uses bounded backoff while considering available quota; treating an outage as a prompt defect misdirects investigation.
Question 557
An organization wants multiple authorized remote clients to connect to a hosted MCP server. What is the most direct implementation choice?
- Return structured errors from tools
- Use a client-side transport adapter
- Use Streamable HTTP for remote MCP access
- Protect remote MCP authentication and authorization
Explanation
Correct Answer
C:A network transport supports remote requests under explicit authentication and deployment controls—with this limitation: opening an HTTP endpoint without access restrictions can expose powerful tools publicly.
Incorrect Answers
A:Explicit error signaling distinguishes tool failure from an empty successful result, whereas the requirement calls for another behavior: a network transport supports remote requests under explicit authentication and deployment controls; swallowing errors can cause a model to hallucinate completion.
B:Transport abstraction lets the same protocol-level concepts operate over supported connection methods; however, the critical mechanism here is that a network transport supports remote requests under explicit authentication and deployment controls, and a transport switch does not erase server deployment or security requirements.
D:Credentials and resource-scoped checks ensure each caller sees only permitted tools and records; this cannot substitute for the behavior that a network transport supports remote requests under explicit authentication and deployment controls, because protocol compatibility alone does not establish who the caller is.
Question 558
A natural-language model proposes a percentage that must be within a strictly allowed range. What is the right implementation response?
- Use retrieval for current factual evidence
- Use a computation tool for exact arithmetic
- Use deterministic validation after generation
- Verify model thinking compatibility
Explanation
Correct Answer
C:Business code checks numeric bounds independently of the model's prose, subject to this requirement: fluent explanations cannot override a failed validation rule.
Incorrect Answers
A:A retrieval step supplies updated source text before answer generation; this cannot substitute for the behavior that business code checks numeric bounds independently of the model's prose, because retrieved content still needs authenticity and relevance checks.
B:A trusted arithmetic implementation calculates the numeric result instead of relying on next-token prediction; in contrast, business code checks numeric bounds independently of the model's prose, while the model can explain the calculation but should not be the authoritative calculator.
D:The integrator checks supported thinking parameters for the chosen version and request mode; here, the controlling capability is that business code checks numeric bounds independently of the model's prose, but feature names and modes can differ between model releases.
Question 559
A newly edited Messages payload is rejected immediately with invalid parameter information. Which solution matches the actual constraint?
- Treat 400-class validation responses as request defects
- Treat HTTP 429 as throttling pressure
- Reproduce with a raw valid API request
- Treat 401/403 responses as credential or access problems
Explanation
Correct Answer
A:The developer checks the request schema, supported fields, and types before attempting another call; the important constraint is that retrying an identical malformed request wastes time and capacity.
Incorrect Answers
B:The caller reduces concurrency and uses bounded backoff while considering available quota; the requirement instead turns on this behavior: the developer checks the request schema, supported fields, and types before attempting another call; blind parallel retries amplify rather than relieve throttle pressure.
C:An isolated direct call distinguishes transport and wrapper behavior from a more complex integration path; this cannot substitute for the behavior that the developer checks the request schema, supported fields, and types before attempting another call, because a successful API response does not automatically validate downstream application logic.
D:Authentication and entitlement checks distinguish invalid credentials from forbidden resource access; in contrast, the developer checks the request schema, supported fields, and types before attempting another call, while changing prompt wording cannot grant missing API permission.
Question 560
A coding agent launched in a repository should use the team's established build and review conventions. Which of the following would work as specified?
- Use CLAUDE.md for repository guidance in Claude Code
- Configure permissions within the relevant client
- Persist conversation history in application storage
- Read project guidance from Claude Code files
Explanation
Correct Answer
D:Claude Code can load contextual project instructions such as CLAUDE.md during its sessions; a crucial qualification is that those instructions do not automatically govern an isolated custom API service.
Incorrect Answers
A:Repository instruction files provide persistent project context to the Claude Code environment; in contrast, claude Code can load contextual project instructions such as CLAUDE.md during its sessions, while they are not automatically loaded into unrelated Messages API clients.
B:Client-specific settings define the environment and its permitted tool behavior; this addresses a separate concern, not the fact that claude Code can load contextual project instructions such as CLAUDE.md during its sessions; configuring a distant backend does not necessarily constrain a local coding session.
C:The application saves and reconstructs relevant turns because Messages requests are stateless; here, the controlling capability is that claude Code can load contextual project instructions such as CLAUDE.md during its sessions, but assuming the API remembers every prior independent call will break continuity.