Question 720
To ship a context-rich document assistant, the application must handle a JSON parse failure without silently accepting partial data and reject model output missing a field that a downstream service requires. Which pair of listed actions is necessary for those stated capabilities? Choose TWO.
- Catch the parsing error and apply a bounded repair or retry path.
- Summarize prior decisions and preserve unresolved constraints.
- Validate parsed output against the application's declared schema.
- Apply domain-specific authorization and business rules after schema validation.
- Have a scoped subagent process sources and return a concise verified handoff.
Explanation
Correct Answers
A:To handle a JSON parse failure without silently accepting partial data as part of a design that must also reject model output missing a field that a downstream service requires, defensive parsing makes failures explicit and limits reattempts; swallowing errors can propagate corrupt records; this is why the action is one of the two required selections.
C:To reject model output missing a field that a downstream service requires as part of a design that must also handle a JSON parse failure without silently accepting partial data, parsing proves a string is legal JSON, but schema validation checks required keys, types, and structural constraints; this is why the action is one of the two required selections.
Incorrect Answers
B:This alternative serves the separate goal to maintain the current task after older dialogue exceeds the active token budget because a faithful rolling summary carries forward objectives and commitments while avoiding a full-history replay on each call; the required outcomes are instead to handle a JSON parse failure without silently accepting partial data and to reject model output missing a field that a downstream service requires.
D:This alternative serves the separate goal to block a technically well-formed refund record with an unauthorized amount because type correctness is not permission to act; amounts, account ownership, and policy limits need independent checks; the required outcomes are instead to handle a JSON parse failure without silently accepting partial data and to reject model output missing a field that a downstream service requires.
E:This alternative serves the separate goal to analyze a large source set without exposing every raw document to the parent agent because separate contexts let specialists work deeply while the coordinator sees only claims and references needed for the final decision; the required outcomes are instead to handle a JSON parse failure without silently accepting partial data and to reject model output missing a field that a downstream service requires.
Question 721
To ship a latency-sensitive model gateway, the application must trade answer latency and computation against solution quality on supported models and explain why language-model output is generated step by step rather than fetched from a fixed response table. Which pair of listed actions is necessary for those stated capabilities? Choose TWO.
- Budget total context for instructions, history, tools, and output.
- Select an available effort level after measuring task performance.
- Model the response as conditioned token generation.
- Evaluate a Sonnet-family model against a realistic task set.
- Use a synchronous client with explicit timeout and error handling.
Explanation
Correct Answers
B:To trade answer latency and computation against solution quality on supported models as part of a design that must also explain why language-model output is generated step by step rather than fetched from a fixed response table, reasoning effort is an inference-time tradeoff; lowering it may reduce costs or latency while harming difficult-case accuracy; this is why the action is one of the two required selections.
C:To explain why language-model output is generated step by step rather than fetched from a fixed response table as part of a design that must also trade answer latency and computation against solution quality on supported models, autoregressive decoding selects subsequent tokens based on supplied context and prior output, which creates flexible but non-guaranteed wording; this is why the action is one of the two required selections.
Incorrect Answers
A:This alternative serves the separate goal to know why a long prompt plus a proposed response may exceed the model's accepted context because a context window bounds what can be processed in one request; large tool results consume it along with user messages; the required outcomes are instead to trade answer latency and computation against solution quality on supported models and to explain why language-model output is generated step by step rather than fetched from a fixed response table.
D:This alternative serves the separate goal to deliver moderately complex development assistance with balanced quality, cost, and latency because a balanced model tier often offers useful capability without the latency or cost of the most expensive tier; evaluation should drive final selection; the required outcomes are instead to trade answer latency and computation against solution quality on supported models and to explain why language-model output is generated step by step rather than fetched from a fixed response table.
E:This option serves the separate goal to implement a low-volume command-line utility that handles one inference at a time because synchronous calls are simplest for strictly sequential work, where concurrency would add complexity without improving throughput; the question instead requires the abilities to trade answer latency and computation against solution quality on supported models and to explain why language-model output is generated step by step rather than fetched from a fixed response table.
Question 722
A change request for a high-volume Messages API gateway combines two needs: limit the blast radius of a new model configuration during rollout, and restore a known-working prompt and model configuration after quality degrades. Which implementation pair addresses both, not just one?
- Perform both steps: (1) Maintain links between requirements, implementation changes, and verification cases; (2) branch on stop_reason tool_use and dispatch the declared calls.
- Perform both steps: (1) Maintain links between requirements, implementation changes, and verification cases; (2) keep versioned artifacts and a tested rollback procedure.
- Perform both steps: (1) Release to a small monitored cohort before broad deployment; (2) keep versioned artifacts and a tested rollback procedure.
- Perform both steps: (1) Release to a small monitored cohort before broad deployment; (2) branch on stop_reason tool_use and dispatch the declared calls.
Explanation
Correct Answer
C:The first change enables the team to limit the blast radius of a new model configuration during rollout because canary exposure provides real traffic signals with an option to stop or roll back, unlike a simultaneous change for every customer; the second addresses the separate need to restore a known-working prompt and model configuration after quality degrades because model IDs, prompts, schemas, and tool contracts are release components; restoring just the application binary may not restore prior behavior.
Incorrect Answers
A:Both steps target different results—ensure each production feature can be tied to an approved acceptance test and recognize that the model is asking the application to execute a client tool—because requirement traceability helps prevent a prompt or model update from silently changing behavior outside the agreed product scope and a tool-use stop requires host-side handling and a follow-up message containing results, not presenting the call as final prose; the required capabilities are instead to limit the blast radius of a new model configuration during rollout and to restore a known-working prompt and model configuration after quality degrades.
B:The second step meets the requirement to restore a known-working prompt and model configuration after quality degrades because model IDs, prompts, schemas, and tool contracts are release components; restoring just the application binary may not restore prior behavior; the first step would only help to ensure each production feature can be tied to an approved acceptance test because requirement traceability helps prevent a prompt or model update from silently changing behavior outside the agreed product scope, leaving the separate requirement to limit the blast radius of a new model configuration during rollout unmet.
D:The first step is appropriate for the requirement to limit the blast radius of a new model configuration during rollout because canary exposure provides real traffic signals with an option to stop or roll back, unlike a simultaneous change for every customer; the other step instead aims to recognize that the model is asking the application to execute a client tool because a tool-use stop requires host-side handling and a follow-up message containing results, not presenting the call as final prose, which does not meet the second requirement to restore a known-working prompt and model configuration after quality degrades.
Question 723
A change request for a tool-using incident triage agent combines two needs: impose predictable maximum cost and wall-clock exposure on autonomous execution, and execute a requested internal business function inside the application's trusted process. Which implementation pair addresses both, not just one?
- Perform both steps: (1) Consume the agent runtime's streamed event sequence; (2) inspect the tool_use block, invoke the authorized handler, and return tool_result.
- Perform both steps: (1) Apply limits for steps, elapsed time, and token or tool expenditure; (2) inspect the tool_use block, invoke the authorized handler, and return tool_result.
- Perform both steps: (1) Apply limits for steps, elapsed time, and token or tool expenditure; (2) persist and resume the agent session using supported state handling.
- Perform both steps: (1) Consume the agent runtime's streamed event sequence; (2) persist and resume the agent session using supported state handling.
Explanation
Correct Answer
B:The first change enables the team to impose predictable maximum cost and wall-clock exposure on autonomous execution because runtime budgets are external constraints on agent activity and must be enforced by the host, not inferred from a cooperative natural-language instruction; the second addresses the separate need to execute a requested internal business function inside the application's trusted process because client tools are executed by the application, not by merely declaring a schema; every returned result must be matched to its tool-use identifier.
Incorrect Answers
A:The second step meets the requirement to execute a requested internal business function inside the application's trusted process because client tools are executed by the application, not by merely declaring a schema; every returned result must be matched to its tool-use identifier; the first step would only help to deliver intermediate agent events to the client as they occur because event-driven consumption lets the application display progress and inspect tool actions without waiting for the entire multi-step run to finish, leaving the separate requirement to impose predictable maximum cost and wall-clock exposure on autonomous execution unmet.
C:The first step is appropriate for the requirement to impose predictable maximum cost and wall-clock exposure on autonomous execution because runtime budgets are external constraints on agent activity and must be enforced by the host, not inferred from a cooperative natural-language instruction; the other step instead aims to continue a paused multi-step task with its legitimate prior context because session continuity preserves relevant work across invocations; discarding session state can cause repeated tools or lost intermediate decisions, which does not meet the second requirement to execute a requested internal business function inside the application's trusted process.
D:Both steps target different results—deliver intermediate agent events to the client as they occur and continue a paused multi-step task with its legitimate prior context—because event-driven consumption lets the application display progress and inspect tool actions without waiting for the entire multi-step run to finish and session continuity preserves relevant work across invocations; discarding session state can cause repeated tools or lost intermediate decisions; the required capabilities are instead to impose predictable maximum cost and wall-clock exposure on autonomous execution and to execute a requested internal business function inside the application's trusted process.
Question 724
A team maintaining a model cost review must meet 3 separate acceptance criteria: (1) identify why verbose generated reports cost more despite unchanged inputs; (2) explain why language-model output is generated step by step rather than fetched from a fixed response table; (3) investigate inconsistent throughput during a traffic spike. Which three implementation actions directly satisfy the stated requirements? Choose THREE.
- Start with a clear instruction and no demonstrations.
- Provide one accurate input-output demonstration.
- Model the response as conditioned token generation.
- Track output tokens separately from input tokens.
- Measure input-token usage per request and track the prompt components.
- Record rate-limit status and relevant response headers alongside request metrics.
Explanation
Correct Answers
C:The test to explain why language-model output is generated step by step rather than fetched from a fixed response table is met because autoregressive decoding selects subsequent tokens based on supplied context and prior output, which creates flexible but non-guaranteed wording; the remaining selections covers the separate need to identify why verbose generated reports cost more despite unchanged inputs and need to investigate inconsistent throughput during a traffic spike.
D:The test to identify why verbose generated reports cost more despite unchanged inputs is met because generation tokens are metered independently and can have different pricing, so response verbosity affects unit economics; the remaining selections covers the separate need to explain why language-model output is generated step by step rather than fetched from a fixed response table and need to investigate inconsistent throughput during a traffic spike.
F:The test to investigate inconsistent throughput during a traffic spike is met because quota-related measurements explain capacity failures that may otherwise look like randomly slow model inference; the remaining selections covers the separate need to explain why language-model output is generated step by step rather than fetched from a fixed response table and need to identify why verbose generated reports cost more despite unchanged inputs.
Incorrect Answers
A:The action is useful for the separate capability to run a straightforward well-specified classification without consuming examples because zero-shot prompting relies on direct task instructions and can save context when the task requires little format teaching; however, the required test outcomes are to explain why language-model output is generated step by step rather than fetched from a fixed response table and to identify why verbose generated reports cost more despite unchanged inputs and to investigate inconsistent throughput during a traffic spike.
B:The action is useful for the separate capability to teach a distinctive output format with one representative example because a single example illustrates structure and conventions, but it may not reveal how to handle difficult edge cases; however, the required test outcomes are to explain why language-model output is generated step by step rather than fetched from a fixed response table and to identify why verbose generated reports cost more despite unchanged inputs and to investigate inconsistent throughput during a traffic spike.
E:The action is useful for the separate capability to explain a billing increase after lengthy system instructions are added because repeated prefixes and conversation history consume billable input, so cost cannot be inferred from output length alone; however, the required test outcomes are to explain why language-model output is generated step by step rather than fetched from a fixed response table and to identify why verbose generated reports cost more despite unchanged inputs and to investigate inconsistent throughput during a traffic spike.
Question 725
A planned upgrade to a multi-tenant support application has two requirements that must both pass: avoid an outage when a pinned model version approaches retirement; limit the blast radius of a new model configuration during rollout. Identify the two direct implementations. Choose TWO.
- Maintain links between requirements, implementation changes, and verification cases.
- Track provider deprecation notices and test migration before the removal date.
- Require structured request IDs, tool traces, and version metadata with privacy controls.
- Treat model migration as a tested deployment with staged traffic.
- Release to a small monitored cohort before broad deployment.
Explanation
Correct Answers
B:To avoid an outage when a pinned model version approaches retirement as part of a design that must also limit the blast radius of a new model configuration during rollout, lifecycle management treats model availability as a dependency and validates output regressions before switching production traffic; this is why the action is one of the two required selections.
E:To limit the blast radius of a new model configuration during rollout as part of a design that must also avoid an outage when a pinned model version approaches retirement, canary exposure provides real traffic signals with an option to stop or roll back, unlike a simultaneous change for every customer; this is why the action is one of the two required selections.
Incorrect Answers
A:This alternative serves the separate goal to ensure each production feature can be tied to an approved acceptance test because requirement traceability helps prevent a prompt or model update from silently changing behavior outside the agreed product scope; the required outcomes are instead to avoid an outage when a pinned model version approaches retirement and to limit the blast radius of a new model configuration during rollout.
C:This alternative serves the separate goal to enable an administrator to determine how a particular AI decision was produced because traceability depends on consistent operational evidence, not just retaining the final generated answer or a screenshot; the required outcomes are instead to avoid an outage when a pinned model version approaches retirement and to limit the blast radius of a new model configuration during rollout.
D:This alternative serves the separate goal to compare candidate models before switching a mission-critical endpoint because model upgrades may alter output distributions or supported features; staged validation reduces the chance of an unnoticed behavior shift; the required outcomes are instead to avoid an outage when a pinned model version approaches retirement and to limit the blast radius of a new model configuration during rollout.
Question 726
A release check on a Python Claude API backend found two unsatisfied conditions: the service must learn from a tool-side effect caused by an unexpected model action, and it must ensure each production feature can be tied to an approved acceptance test. Which complete engineering plan satisfies both?
- Perform both steps: (1) Store prompt versions with release identifiers and rollback history; (2) branch on stop_reason tool_use and dispatch the declared calls.
- Perform both steps: (1) Record a blameless timeline, root causes, and preventive controls; (2) branch on stop_reason tool_use and dispatch the declared calls.
- Perform both steps: (1) Store prompt versions with release identifiers and rollback history; (2) maintain links between requirements, implementation changes, and verification cases.
- Perform both steps: (1) Record a blameless timeline, root causes, and preventive controls; (2) maintain links between requirements, implementation changes, and verification cases.
Explanation
Correct Answer
D:The first change enables the team to learn from a tool-side effect caused by an unexpected model action because post-incident review should identify inadequate approvals and monitoring and turn findings into testable changes, rather than blaming the language model; the second addresses the separate need to ensure each production feature can be tied to an approved acceptance test because requirement traceability helps prevent a prompt or model update from silently changing behavior outside the agreed product scope.
Incorrect Answers
A:Both steps target different results—determine whether a quality regression came from a recent instruction edit and recognize that the model is asking the application to execute a client tool—because a prompt is a behavior-changing artifact; versioning makes comparisons and safe reversions possible during incident response and a tool-use stop requires host-side handling and a follow-up message containing results, not presenting the call as final prose; the required capabilities are instead to learn from a tool-side effect caused by an unexpected model action and to ensure each production feature can be tied to an approved acceptance test.
B:The first step is appropriate for the requirement to learn from a tool-side effect caused by an unexpected model action because post-incident review should identify inadequate approvals and monitoring and turn findings into testable changes, rather than blaming the language model; the other step instead aims to recognize that the model is asking the application to execute a client tool because a tool-use stop requires host-side handling and a follow-up message containing results, not presenting the call as final prose, which does not meet the second requirement to ensure each production feature can be tied to an approved acceptance test.
C:The second step meets the requirement to ensure each production feature can be tied to an approved acceptance test because requirement traceability helps prevent a prompt or model update from silently changing behavior outside the agreed product scope; the first step would only help to determine whether a quality regression came from a recent instruction edit because a prompt is a behavior-changing artifact; versioning makes comparisons and safe reversions possible during incident response, leaving the separate requirement to learn from a tool-side effect caused by an unexpected model action unmet.
Question 727
The team responsible for a latency-sensitive model gateway needs to process partial model output as it is delivered over a streaming connection. What is the most direct engineering change?
- Design for stochastic output and evaluate behavior over representative repeats.
- Iterate the SDK or HTTP event stream until its terminal event.
- Serialize the documented nested request shape rather than flattening it.
- Branch on HTTP status and decode the provider's error structure.
Explanation
Correct Answer
B:To process partial model output as it is delivered over a streaming connection in this latency-sensitive model gateway, streaming clients must handle both content deltas and final status rather than assuming the first chunk is the complete response; the proposed implementation directly supplies that capability.
Incorrect Answers
A:The proposed alternative helps to account for different plausible answers from the same input on repeated runs because sampling and model execution can produce variation; important correctness must be checked rather than assumed identical on every call; that mechanism does not meet the different requirement to process partial model output as it is delivered over a streaming connection.
C:The proposed alternative helps to preserve arrays of typed content blocks when sending multimodal input because typed content and role fields carry protocol meaning that is lost if developers simply concatenate them into plain text; that mechanism does not meet the different requirement to process partial model output as it is delivered over a streaming connection.
D:The proposed alternative helps to stop interpreting non-2xx API responses as normal assistant messages because transport errors are not regular model completions, so application control flow must treat them separately; that mechanism does not meet the different requirement to process partial model output as it is delivered over a streaming connection.
Question 728
A test for a production inference application fails because it cannot send a model inference request with the operation expected by the REST endpoint. Which change fixes the identified failure?
- Use the API's documented POST request with a JSON body.
- Give mutating operations an application-level idempotency key.
- Place a valid secret in the documented API authentication header.
- Reuse an appropriately configured HTTP client or SDK connection pool.
Explanation
Correct Answer
A:To send a model inference request with the operation expected by the REST endpoint in this production inference application, a state-changing inference operation is submitted as an HTTP request with an explicit method and payload, not inferred from a browser URL alone; the proposed implementation directly supplies that capability.
Incorrect Answers
B:The proposed alternative helps to prevent duplicate purchases or tickets if an agent repeats a tool call because idempotency ensures replays of the same intended action do not create duplicate real-world effects even when orchestration retries; that mechanism does not meet the different requirement to send a model inference request with the operation expected by the REST endpoint.
C:The proposed alternative helps to authenticate a server-to-server Claude API request because API authentication is a request transport concern; placing a key inside the prompt body neither authenticates the request nor protects it; that mechanism does not meet the different requirement to send a model inference request with the operation expected by the REST endpoint.
D:The proposed alternative helps to reduce repeated TLS setup overhead for high-volume application traffic because connection reuse lowers transport overhead but still requires limits and health handling for long-lived clients; that mechanism does not meet the different requirement to send a model inference request with the operation expected by the REST endpoint.
Question 729
Two approval criteria are still open for a business workflow that calls Claude: the team must map out-of-order batch outputs back to source records reliably and connect execution output to the exact tool request Claude made. Which two implementation actions satisfy them? Choose TWO.
- Configure an available server-side tool with the required API definition.
- Return tool_result blocks containing the corresponding tool_use_id.
- Use the provider-specific endpoint, authentication, and request conventions.
- Send instructions using the request's top-level system field.
- Assign and preserve unique custom identifiers for batched requests.
Explanation
Correct Answers
B:To connect execution output to the exact tool request Claude made as part of a design that must also map out-of-order batch outputs back to source records reliably, tool correlation preserves call-to-result semantics in the message sequence, especially when several tools were requested together; this is why the action is one of the two required selections.
E:To map out-of-order batch outputs back to source records reliably as part of a design that must also connect execution output to the exact tool request Claude made, asynchronous results need stable correlation keys; relying on processing order can attach an output to the wrong input; this is why the action is one of the two required selections.
Incorrect Answers
A:This alternative serves the separate goal to use provider-hosted search or execution without implementing the tool body locally because provider-executed tools are handled in Anthropic's environment and differ from client functions that the caller must run; the required outcomes are instead to map out-of-order batch outputs back to source records reliably and to connect execution output to the exact tool request Claude made.
C:This alternative serves the separate goal to support Claude models accessed through different supported cloud offerings because Claude API, Bedrock, and Vertex-style integrations are not interchangeable URLs; host authentication and service wrappers differ; the required outcomes are instead to map out-of-order batch outputs back to source records reliably and to connect execution output to the exact tool request Claude made.
D:This alternative serves the separate goal to set consistent developer-controlled behavior separately from untrusted user text because the system field separates higher-priority application guidance from user-message content and should not be simulated by ordinary quoted text; the required outcomes are instead to map out-of-order batch outputs back to source records reliably and to connect execution output to the exact tool request Claude made.
Question 730
For a high-volume Messages API gateway, the missing capability is to authenticate a server-to-server Claude API request. What should the developers implement?
- Place a valid secret in the documented API authentication header.
- Require peer review focused on API, security, and tool-side effects.
- Add automated consumer-provider contract tests.
- Retry eligible failures with capped exponential backoff and jitter.
Explanation
Correct Answer
A:To authenticate a server-to-server Claude API request in this high-volume Messages API gateway, API authentication is a request transport concern; placing a key inside the prompt body neither authenticates the request nor protects it; the proposed implementation directly supplies that capability.
Incorrect Answers
B:The proposed alternative helps to catch unsafe credentials and incorrect tool calls before deployment because human review should check trust boundaries and external actions in addition to style; successful compilation cannot prove these behaviors are safe; that mechanism does not meet the different requirement to authenticate a server-to-server Claude API request.
C:The proposed alternative helps to detect a changed tool response field before it breaks downstream code because schema contracts make integration changes visible and prevent a renamed field from silently corrupting later model context; that mechanism does not meet the different requirement to authenticate a server-to-server Claude API request.
D:The proposed alternative helps to recover transient 429 or 5xx responses without creating a retry storm because backoff spreads retry traffic and respects transient capacity limits; blind immediate repetition worsens overload; that mechanism does not meet the different requirement to authenticate a server-to-server Claude API request.