Question 80
A support agent uses evolving retrieval content, but developers only test the original prompt. What lifecycle risk remains?
- Review versioned prompts on representative regression cases before deployment.
- Monitor data and knowledge-base changes as inputs to regression evaluation.
- Offer rollout controls and a fast feature rollback or kill switch.
- Set measurable acceptance criteria and responsibility for ongoing quality.
Explanation
Correct Answer
B:A stable prompt does not guarantee stable results when source documents, schemas, and retrieval rankings change.
Incorrect Answers
A:A prompt is executable application behavior; test gates and rollback history should cover it like code; for data-driven regression drift, a control centered on prompt release regression testing is not the appropriate substitute.
C:Operational containment limits impact when a new model, prompt, or connector misbehaves after deployment; this implements controllable production rollback rather than the demanded data-driven regression drift.
D:AI capabilities require defined ownership for evaluations, observed failures, and user-facing policy decisions after release; testing data-driven regression drift would still fail after improving lifecycle ownership.
Question 81
A team refactors the Claude integration while preserving its public contract. What checks are most relevant?
- Perform integration tests with simulated throttling and dependency failures.
- Specify and rehearse restoration of configuration, durable state, and secret references.
- Provide runtime observability and incident response for model and tool outcomes.
- Introduce privacy transformations to design, implementation, and automated verification.
- Compare API contract tests and representative behavior before and after the change.
Explanation
Correct Answer
E:Internal restructuring should not silently alter externally promised response schemas, error behavior, or supported user workflows.
Incorrect Answers
A:Unit tests of isolated logic do not prove the complete request, retry, and tool-response path behaves correctly; the scenario's controlling requirement is refactoring contract preservation, not the related integration failure testing concern.
B:Recovery requires more than redeploying code; operations must reconstruct approved settings and resume tasks without duplicating side effects; the consequence is improved operational recovery readiness, not verified refactoring contract preservation.
C:Deployment is not the end of the system lifecycle; ongoing traces, metrics, and triage are needed as inputs and dependencies change; where refactoring contract preservation matters, a control for postdeployment operations leaves that test unmet.
D:Compliance is cross-cutting; a late manual checklist cannot reliably protect every request path; the effect concerns privacy controls across SDLC; the success condition is refactoring contract preservation.
Question 82
A feature request describes a useful demo but lacks a success owner. What should product delivery establish?
- Assess API contract tests and representative behavior before and after the change.
- Pin and test integration contracts, and handle incompatible payloads explicitly.
- Assign measurable acceptance criteria and responsibility for ongoing quality.
- Test versioned prompts on representative regression cases before deployment.
Explanation
Correct Answer
C:AI capabilities require defined ownership for evaluations, observed failures, and user-facing policy decisions after release.
Incorrect Answers
A:Internal restructuring should not silently alter externally promised response schemas, error behavior, or supported user workflows; this may help refactoring contract preservation but cannot by itself deliver lifecycle ownership.
B:Schema evolution affects the application's tool boundary even if model code is untouched; contract discipline catches breaks early; where lifecycle ownership matters, a control for dependent API schema evolution leaves that test unmet.
D:A prompt is executable application behavior; test gates and rollback history should cover it like code; for lifecycle ownership, this instead manages prompt release regression testing; the requested decision remains unresolved.
Question 83
A third-party tool changes its response JSON field names without warning. What engineering process reduces risk?
- Insert runtime observability and incident response for model and tool outcomes.
- Give rollout controls and a fast feature rollback or kill switch.
- Version and test integration contracts, and handle incompatible payloads explicitly.
- Provide privacy transformations to design, implementation, and automated verification.
Explanation
Correct Answer
C:Schema evolution affects the application's tool boundary even if model code is untouched; contract discipline catches breaks early.
Incorrect Answers
A:Deployment is not the end of the system lifecycle; ongoing traces, metrics, and triage are needed as inputs and dependencies change; the dependent API schema evolution criterion differs from postdeployment operations; the gap persists.
B:Operational containment limits impact when a new model, prompt, or connector misbehaves after deployment; when the goal is dependent API schema evolution, optimizing controllable production rollback misses that operational boundary.
D:Compliance is cross-cutting; a late manual checklist cannot reliably protect every request path; even if privacy controls across SDLC succeeds, dependent API schema evolution still needs separate verification.
Question 84
A Claude application needs repeatable disaster recovery after losing its orchestrator host. Which lifecycle deliverable matters?
- Deliver rollout controls and a fast feature rollback or kill switch.
- Insert privacy transformations to design, implementation, and automated verification.
- Designate measurable acceptance criteria and responsibility for ongoing quality.
- Document and rehearse restoration of configuration, durable state, and secret references.
Explanation
Correct Answer
D:Recovery requires more than redeploying code; operations must reconstruct approved settings and resume tasks without duplicating side effects.
Incorrect Answers
A:Operational containment limits impact when a new model, prompt, or connector misbehaves after deployment; the scenario's controlling requirement is operational recovery readiness, not the related controllable production rollback concern.
B:Compliance is cross-cutting; a late manual checklist cannot reliably protect every request path; where operational recovery readiness matters, a control for privacy controls across SDLC leaves that test unmet.
C:AI capabilities require defined ownership for evaluations, observed failures, and user-facing policy decisions after release; for operational recovery readiness, a control centered on lifecycle ownership is not the appropriate substitute.
Question 85
An HTTP client posts to the Messages API without specifying a token limit. Which request field should the developer include?
- Supply a valid max_tokens output budget for the message.
- Employ the top-level system field for application-level instructions.
- Continue or retry with sufficient generation allowance after assessing partial output.
- Provide the image as a supported image content block in a user message.
Explanation
Correct Answer
A:The Messages request contract includes an output limit; omitting required generation parameters causes an invalid request rather than unlimited output.
Incorrect Answers
B:The API has a distinct system instruction surface; role confusion weakens the separation between application policy and user content; this targets system instruction field, while the system must demonstrate required output budget.
C:The response ended at the configured limit rather than a natural completion; treating it as final risks publishing an incomplete answer; this implements generation limit stop rather than the demanded required output budget.
D:Vision is driven by multimodal message content, not by giving Claude a local filename it cannot access; for required output budget, this instead manages API image content structure; the requested decision remains unresolved.
Question 86
A Python service sends a multi-turn conversation but assumes Claude retains its previous requests on the server. What must be corrected?
- Resend the relevant prior messages or maintained conversation state with each request.
- Examine the usage fields returned with the API response.
- Process content blocks by type rather than reading only the first text item.
- Send independent Messages requests through the Message Batches API.
Explanation
Correct Answer
A:The Messages API is stateless with respect to caller history; the client owns the turns it wants Claude to use.
Incorrect Answers
B:Request and response token accounting is exposed as metadata; text length alone is an unreliable measure of billed tokens; for this request, explicit conversation history is the missing piece, not response token usage metadata.
C:Messages responses can be heterogeneous; ignoring block types can hide tool requests or other structured content; it changes heterogeneous response blocks; it does not prove explicit conversation history in this case.
D:Asynchronous batches suit throughput-oriented, latency-tolerant jobs; they are not a synchronous streaming response channel; the consequence is improved asynchronous independent classification, not verified explicit conversation history.
Question 87
An engineer places the assistant's governing instructions as a fabricated user turn. What is the correct API design?
- Provide a valid max_tokens output budget for the message.
- Examine the permitted file and attach supported content through the API.
- Stop before executing the mutating client tool and await explicit approval.
- Use the top-level system field for application-level instructions.
Explanation
Correct Answer
D:The API has a distinct system instruction surface; role confusion weakens the separation between application policy and user content.
Incorrect Answers
A:The Messages request contract includes an output limit; omitting required generation parameters causes an invalid request rather than unlimited output; even if required output budget succeeds, system instruction field still needs separate verification.
B:A pathname is not a transferable attachment; the caller controls file access, serialization, and privacy checks before a request is sent; even if local file ingestion by caller succeeds, system instruction field still needs separate verification.
C:The model may generate a valid tool request, but the host owns side effects and can withhold execution until policy authorization is satisfied; for system instruction field, this instead manages approval before tool execution; the requested decision remains unresolved.
Question 88
A custom REST client receives a 401 while its SDK-based test succeeds. What should it inspect first?
- Examine tool_choice, tool definitions, and the returned stop_reason and content blocks.
- Shift changing context after the stable cached prefix.
- Verify that the client sends the required authentication and API-version headers.
- Turn on prompt caching on the reusable prefix of the request.
Explanation
Correct Answer
C:A raw HTTP integration must provide the headers the SDK normally adds; missing credentials or version metadata prevents the request from being accepted.
Incorrect Answers
A:Tool invocation can be model-selected or forced within supported modes; observing the actual response distinguishes selection behavior from integration failure; the REST request headers criterion differs from tool selection diagnostics; the gap persists.
B:Cache keys depend on the preceding content and ordering, so early variation invalidates reuse even when the policy text is unchanged; testing REST request headers would still fail after improving cache prefix stability.
D:Cache hits avoid reprocessing the identical policy prefix; dynamic user content should appear after the stable portion; it changes shared request prefix reuse; it does not prove REST request headers in this case.
Question 89
A developer needs to show a response token by token as it arrives. Which API pattern meets that presentation requirement?
- Request streaming and consume server-sent events incrementally.
- Employ the provider-specific endpoint, credentials, and supported feature surface.
- Employ a document input path supporting visual PDF processing on the chosen platform.
- Examine refusal or stop state in the model response separately from HTTP transport success.
- Confirm that the client sends the required authentication and API-version headers.
Explanation
Correct Answer
A:Streaming yields content deltas before the complete message is assembled, reducing perceived wait time for the user.
Incorrect Answers
B:Bedrock or another host may wrap Claude with distinct API conventions and feature availability; native API assumptions do not automatically transfer; it changes third-party provider integration; it does not prove incremental text presentation in this case.
C:Extraction of plain text alone can omit layout and charts; document support and provider-specific feature parity matter; it improves visually rich PDF ingestion only, leaving incremental text presentation unproven.
D:Successful delivery of a response does not imply the content fulfills the task; refusal is an application-level outcome; a remedy for model refusal versus network success is not a substitute for incremental text presentation.
E:A raw HTTP integration must provide the headers the SDK normally adds; missing credentials or version metadata prevents the request from being accepted; a remedy for REST request headers is not a substitute for incremental text presentation.
Question 90
An API stream occasionally ends after receiving partial content. What must the consumer not assume?
- Activate prompt caching on the reusable prefix of the request.
- Do not treat partial streaming deltas as a finalized, validated response.
- Use supported asynchronous SDK calls or async HTTP I/O with bounded concurrency.
- Process the intermediate tool round trip before rendering a completed assistant turn.
Explanation
Correct Answer
B:A stream must be assembled and its final completion or error state inspected; interim bytes may not form a complete answer.
Incorrect Answers
A:Cache hits avoid reprocessing the identical policy prefix; dynamic user content should appear after the stable portion; this may help shared request prefix reuse but cannot by itself deliver streaming completion safety.
C:Nonblocking request handling lets the service overlap remote latency while maintaining explicit limits on in-flight work; for streaming completion safety, this instead manages multi-criterion integration; the requested decision remains unresolved.
D:Tool-enabled interactions can require multiple Messages calls; the final user experience is not always one HTTP request; this targets multi-call conversational tool flow, while the system must demonstrate streaming completion safety.