Question 560
A test for a multi-stage support automation fails because it cannot prevent a research assistant from unexpectedly making production changes. Which change fixes the identified failure?
- Give the research subagent read-only tools and a narrow instruction contract.
- Run side-effect-free tool calls concurrently when their inputs are independent.
- Catch tool failures and report typed errors into the agent loop.
- Use isolated branch state and a controlled reconciliation step.
Explanation
Correct Answer
A:To prevent a research assistant from unexpectedly making production changes in this multi-stage support automation, capability isolation prevents a specialist from acting outside its delegated role even if it encounters persuasive untrusted instructions; the proposed implementation directly supplies that capability.
Incorrect Answers
B:The proposed alternative helps to speed up independent data lookups without causing write-order races because read-only independent calls can safely overlap; mutating operations need explicit ordering or concurrency control; that mechanism does not meet the different requirement to prevent a research assistant from unexpectedly making production changes.
C:The proposed alternative helps to keep a failed tool from being mistaken for successful completion of an objective because explicit failure messages let the model or controller choose recovery; swallowing exceptions makes later actions rely on results that never existed; that mechanism does not meet the different requirement to prevent a research assistant from unexpectedly making production changes.
D:The proposed alternative helps to stop two parallel branches from overwriting the same mutable record because explicit merge rules are required when outputs converge; concurrent mutation without arbitration causes nondeterministic results; that mechanism does not meet the different requirement to prevent a research assistant from unexpectedly making production changes.
Question 561
A release check on an enterprise agent deployment found two unsatisfied conditions: the service must identify prohibited requests before an automated agent invokes business tools, and it must require a person to approve an irreversible administrative change. Which complete engineering plan satisfies both?
- Perform both steps: (1) Apply an input-side policy gate and route flagged cases to a safe path; (2) redact secrets and protect log access with retention controls.
- Perform both steps: (1) Apply an input-side policy gate and route flagged cases to a safe path; (2) pause execution for explicit approval at the high-impact action boundary.
- Perform both steps: (1) Verify origin and integrity of sensitive external data where supported; (2) pause execution for explicit approval at the high-impact action boundary.
- Perform both steps: (1) Verify origin and integrity of sensitive external data where supported; (2) redact secrets and protect log access with retention controls.
Explanation
Correct Answer
B:The first change enables the team to identify prohibited requests before an automated agent invokes business tools because early filtering reduces inappropriate work, but later checks remain necessary because tool outputs can also introduce risks; the second addresses the separate need to require a person to approve an irreversible administrative change because human approval is a policy-enforced transition, not a note written in the model's plan or a message displayed afterward.
Incorrect Answers
A:The first step is appropriate for the requirement to identify prohibited requests before an automated agent invokes business tools because early filtering reduces inappropriate work, but later checks remain necessary because tool outputs can also introduce risks; the other step instead aims to investigate user actions without leaking API keys into operational records because logs support accountability only when access and retention are restricted; copying all prompt content into logs creates another exposure surface, which does not meet the second requirement to require a person to approve an irreversible administrative change.
C:The second step meets the requirement to require a person to approve an irreversible administrative change because human approval is a policy-enforced transition, not a note written in the model's plan or a message displayed afterward; the first step would only help to detect unauthorized modifications to a tool result before it reaches the model because trusted decisions require confidence that tool outputs were not altered; transport and signature checks complement content validation, leaving the separate requirement to identify prohibited requests before an automated agent invokes business tools unmet.
D:Both steps target different results—detect unauthorized modifications to a tool result before it reaches the model and investigate user actions without leaking API keys into operational records—because trusted decisions require confidence that tool outputs were not altered; transport and signature checks complement content validation and logs support accountability only when access and retention are restricted; copying all prompt content into logs creates another exposure surface; the required capabilities are instead to identify prohibited requests before an automated agent invokes business tools and to require a person to approve an irreversible administrative change.
Question 562
For a model-family migration plan, the missing capability is to avoid overlooking slow tail requests hidden by a good average. What should the developers implement?
- Alert on unusual per-job token and tool-call growth.
- Track output tokens separately from input tokens.
- Tune sampling for lower variance while retaining output validation.
- Report p50, p95, and p99 latency by model and operation.
Explanation
Correct Answer
D:To avoid overlooking slow tail requests hidden by a good average in this model-family migration plan, percentiles expose user experience at the tail and can reveal bottlenecks that a mean response time conceals; the proposed implementation directly supplies that capability.
Incorrect Answers
A:The proposed alternative helps to detect a looping agent that suddenly multiplies token consumption because monitoring consumption across an entire agent run catches runaway workflows that per-request logs may not reveal; that mechanism does not meet the different requirement to avoid overlooking slow tail requests hidden by a good average.
B:The proposed alternative helps to identify why verbose generated reports cost more despite unchanged inputs because generation tokens are metered independently and can have different pricing, so response verbosity affects unit economics; that mechanism does not meet the different requirement to avoid overlooking slow tail requests hidden by a good average.
C:The proposed alternative helps to favor more consistent word choices in a constrained transformation task when supported because temperature influences token sampling; it cannot by itself guarantee truth, policy compliance, or reproducible structured output; that mechanism does not meet the different requirement to avoid overlooking slow tail requests hidden by a good average.
Question 563
A design review for a tool-using incident triage agent asks for support to impose predictable maximum cost and wall-clock exposure on autonomous execution, together with a way to inspect the result of a completed tool call before later model reasoning consumes it. Select the two appropriate changes. Choose TWO.
- Configure tool permissions rather than depending only on a prompt.
- Inspect the tool_use block, invoke the authorized handler, and return tool_result.
- Delegate source analysis to a subagent and return a concise result.
- Attach a post-execution observation hook to the tool lifecycle.
- Apply limits for steps, elapsed time, and token or tool expenditure.
Explanation
Correct Answers
D:To inspect the result of a completed tool call before later model reasoning consumes it as part of a design that must also impose predictable maximum cost and wall-clock exposure on autonomous execution, post-tool inspection can record or validate outcomes after execution, but it cannot retroactively prevent side effects from the completed call; this is why the action is one of the two required selections.
E:To impose predictable maximum cost and wall-clock exposure on autonomous execution as part of a design that must also inspect the result of a completed tool call before later model reasoning consumes it, runtime budgets are external constraints on agent activity and must be enforced by the host, not inferred from a cooperative natural-language instruction; this is why the action is one of the two required selections.
Incorrect Answers
A:This alternative serves the separate goal to restrict a development agent to a defined set of file and command capabilities because runtime permissions constrain what actions can run even if the model requests more access; prompt instructions alone are not a security boundary; the required outcomes are instead to impose predictable maximum cost and wall-clock exposure on autonomous execution and to inspect the result of a completed tool call before later model reasoning consumes it.
B:This alternative serves the separate goal 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 required outcomes are instead to impose predictable maximum cost and wall-clock exposure on autonomous execution and to inspect the result of a completed tool call before later model reasoning consumes it.
C:This alternative serves the separate goal to keep a large source-reading task from crowding the coordinator's working context because isolated workspaces let a specialist process extensive material while the parent receives only task-relevant evidence instead of every intermediate token; the required outcomes are instead to impose predictable maximum cost and wall-clock exposure on autonomous execution and to inspect the result of a completed tool call before later model reasoning consumes it.
Question 564
In a TypeScript Claude integration, a release requirement is to use provider-hosted search or execution without implementing the tool body locally. Which implementation addresses that specific requirement?
- Branch on stop_reason tool_use and dispatch the declared calls.
- Parse response content by block type and follow current thinking-field rules.
- Configure an available server-side tool with the required API definition.
- Use the provider-specific endpoint, authentication, and request conventions.
Explanation
Correct Answer
C:To use provider-hosted search or execution without implementing the tool body locally in this TypeScript Claude integration, provider-executed tools are handled in Anthropic's environment and differ from client functions that the caller must run; the proposed implementation directly supplies that capability.
Incorrect Answers
A:The proposed alternative helps 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; that mechanism does not meet the different requirement to use provider-hosted search or execution without implementing the tool body locally.
B:The proposed alternative helps to handle a response containing reasoning-related blocks without confusing them with ordinary final text because structured content may contain distinct block types and signatures; a parser that assumes only plain text can discard important protocol data; that mechanism does not meet the different requirement to use provider-hosted search or execution without implementing the tool body locally.
D:The proposed alternative helps 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; that mechanism does not meet the different requirement to use provider-hosted search or execution without implementing the tool body locally.
Question 565
The technical owner of a structured output service has to distinguish a user's instructions from a quoted document containing imperatives. A separate requirement is to avoid a prohibited output element such as confidential identifiers. Which two-part remediation covers both?
- Perform both steps: (1) Have a scoped subagent process sources and return a concise verified handoff; (2) specify the forbidden content alongside an enforceable validation check.
- Perform both steps: (1) Mark untrusted quoted material with explicit boundaries and role context; (2) ground consequential statements in retrieved evidence or external checks.
- Perform both steps: (1) Mark untrusted quoted material with explicit boundaries and role context; (2) specify the forbidden content alongside an enforceable validation check.
- Perform both steps: (1) Have a scoped subagent process sources and return a concise verified handoff; (2) ground consequential statements in retrieved evidence or external checks.
Explanation
Correct Answer
C:The first change enables the team to distinguish a user's instructions from a quoted document containing imperatives because delimiters clarify that imported content is data to analyze, not a new authority to change tools or objectives; the second addresses the separate need to avoid a prohibited output element such as confidential identifiers because negative prompt guidance signals the intended boundary, but a deterministic validator is needed where violating it has consequences.
Incorrect Answers
A:The second step meets the requirement to avoid a prohibited output element such as confidential identifiers because negative prompt guidance signals the intended boundary, but a deterministic validator is needed where violating it has consequences; the first step would only help 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, leaving the separate requirement to distinguish a user's instructions from a quoted document containing imperatives unmet.
B:The first step is appropriate for the requirement to distinguish a user's instructions from a quoted document containing imperatives because delimiters clarify that imported content is data to analyze, not a new authority to change tools or objectives; the other step instead aims to avoid treating fluent model wording as evidence that a claim is verified because linguistic confidence is not a calibrated truth guarantee, so factual correctness requires independent support, which does not meet the second requirement to avoid a prohibited output element such as confidential identifiers.
D:Both steps target different results—analyze a large source set without exposing every raw document to the parent agent and avoid treating fluent model wording as evidence that a claim is verified—because separate contexts let specialists work deeply while the coordinator sees only claims and references needed for the final decision and linguistic confidence is not a calibrated truth guarantee, so factual correctness requires independent support; the required capabilities are instead to distinguish a user's instructions from a quoted document containing imperatives and to avoid a prohibited output element such as confidential identifiers.
Question 566
An architecture decision for an MCP integration server must support expose an MCP server to clients across a network boundary and make reference material available for clients to read as context. Which option gives a complete rather than partial solution?
- Perform both steps: (1) Configure a supported remote HTTP transport with authentication and deployment controls; (2) host a restricted client tool in the private application environment.
- Perform both steps: (1) Register a prompt template with its required arguments; (2) expose the material as an mcp resource with a stable identifier.
- Perform both steps: (1) Register a prompt template with its required arguments; (2) host a restricted client tool in the private application environment.
- Perform both steps: (1) Configure a supported remote HTTP transport with authentication and deployment controls; (2) expose the material as an mcp resource with a stable identifier.
Explanation
Correct Answer
D:The first change enables the team to expose an MCP server to clients across a network boundary because remote connectivity needs network routing, identity, and authorization; a local stdio subprocess is not a remotely reachable service; the second addresses the separate need to make reference material available for clients to read as context because resources represent consumable data, contrasting with tools that invoke an operation or prompts that package reusable instructions.
Incorrect Answers
A:The first step is appropriate for the requirement to expose an MCP server to clients across a network boundary because remote connectivity needs network routing, identity, and authorization; a local stdio subprocess is not a remotely reachable service; the other step instead aims to execute a high-privilege enterprise action that must remain inside a corporate network because a private handler provides local network and IAM controls that a generic provider-executed server tool may not possess, which does not meet the second requirement to make reference material available for clients to read as context.
B:The second step meets the requirement to make reference material available for clients to read as context because resources represent consumable data, contrasting with tools that invoke an operation or prompts that package reusable instructions; the first step would only help to offer users a reusable parameterized instruction template from an MCP server because MCP prompts provide structured prompt content, but by themselves do not perform external business actions, leaving the separate requirement to expose an MCP server to clients across a network boundary unmet.
C:Both steps target different results—offer users a reusable parameterized instruction template from an MCP server and execute a high-privilege enterprise action that must remain inside a corporate network—because MCP prompts provide structured prompt content, but by themselves do not perform external business actions and a private handler provides local network and IAM controls that a generic provider-executed server tool may not possess; the required capabilities are instead to expose an MCP server to clients across a network boundary and to make reference material available for clients to read as context.
Question 567
A change request for a Claude API workload benchmark combines two needs: avoid assuming a model-version change preserves tool output and writing behavior, and serve high-volume straightforward classification at a tight latency and cost target. Which implementation pair addresses both, not just one?
- Perform both steps: (1) Revalidate prompts, schemas, safety constraints, and tools after model migration; (2) benchmark a suitable haiku-family model for the simple task.
- Perform both steps: (1) Revalidate prompts, schemas, safety constraints, and tools after model migration; (2) route simple requests to a cheaper evaluated tier and escalate difficult ones.
- Perform both steps: (1) Benchmark an Opus-family model on the hardest cases; (2) route simple requests to a cheaper evaluated tier and escalate difficult ones.
- Perform both steps: (1) Benchmark an Opus-family model on the hardest cases; (2) benchmark a suitable haiku-family model for the simple task.
Explanation
Correct Answer
A:The first change enables the team to avoid assuming a model-version change preserves tool output and writing behavior because new releases can change response distributions or supported behaviors, making a model upgrade a real software change; the second addresses the separate need to serve high-volume straightforward classification at a tight latency and cost target because lightweight model tiers can favor speed and unit economics, but accuracy must still meet the specific evaluation threshold.
Incorrect Answers
B:The first step is appropriate for the requirement to avoid assuming a model-version change preserves tool output and writing behavior because new releases can change response distributions or supported behaviors, making a model upgrade a real software change; the other step instead aims to minimize average model cost when most requests are easy but some require deeper reasoning because cascading matches compute to task difficulty but requires a reliable escalation rule and measurement of misrouting, which does not meet the second requirement to serve high-volume straightforward classification at a tight latency and cost target.
C:Both steps target different results—solve demanding multistep reasoning tasks where quality matters more than lowest unit price and minimize average model cost when most requests are easy but some require deeper reasoning—because higher-capability tiers may improve difficult-task performance, but their cost and latency still need objective justification and cascading matches compute to task difficulty but requires a reliable escalation rule and measurement of misrouting; the required capabilities are instead to avoid assuming a model-version change preserves tool output and writing behavior and to serve high-volume straightforward classification at a tight latency and cost target.
D:The second step meets the requirement to serve high-volume straightforward classification at a tight latency and cost target because lightweight model tiers can favor speed and unit economics, but accuracy must still meet the specific evaluation threshold; the first step would only help to solve demanding multistep reasoning tasks where quality matters more than lowest unit price because higher-capability tiers may improve difficult-task performance, but their cost and latency still need objective justification, leaving the separate requirement to avoid assuming a model-version change preserves tool output and writing behavior unmet.
Question 568
To ship a model-output validation layer, the application must analyze a large source set without exposing every raw document to the parent agent and identify when a long agent discussion begins solving a different problem than the user approved. Which pair of listed actions is necessary for those stated capabilities? Choose TWO.
- Recheck current plans against a stored objective and constraints.
- Have a scoped subagent process sources and return a concise verified handoff.
- Put stable application constraints in the system instruction field.
- Minimize or redact sensitive fields before constructing context.
- Mark untrusted quoted material with explicit boundaries and role context.
Explanation
Correct Answers
A:To identify when a long agent discussion begins solving a different problem than the user approved as part of a design that must also analyze a large source set without exposing every raw document to the parent agent, explicit objective comparison detects drift before tool actions diverge from the original task; extra history alone cannot guarantee alignment; this is why the action is one of the two required selections.
B:To analyze a large source set without exposing every raw document to the parent agent as part of a design that must also identify when a long agent discussion begins solving a different problem than the user approved, separate contexts let specialists work deeply while the coordinator sees only claims and references needed for the final decision; this is why the action is one of the two required selections.
Incorrect Answers
C:This alternative serves the separate goal to apply the same compliance restrictions across many different end-user requests because high-priority application instructions should not be delivered as ordinary user content where they compete with user questions and document text; the required outcomes are instead to analyze a large source set without exposing every raw document to the parent agent and to identify when a long agent discussion begins solving a different problem than the user approved.
D:This alternative serves the separate goal to provide a document's necessary meaning without disclosing irrelevant personal identifiers because data minimization reduces the sensitive information transmitted for inference while retaining the facts the task legitimately requires; the required outcomes are instead to analyze a large source set without exposing every raw document to the parent agent and to identify when a long agent discussion begins solving a different problem than the user approved.
E:This alternative serves the separate goal to distinguish a user's instructions from a quoted document containing imperatives because delimiters clarify that imported content is data to analyze, not a new authority to change tools or objectives; the required outcomes are instead to analyze a large source set without exposing every raw document to the parent agent and to identify when a long agent discussion begins solving a different problem than the user approved.
Question 569
While planning a function-calling service, the product owner specifies that it must reduce context bloat caused by a tool returning a thousand irrelevant fields. Which solution fulfills the requirement?
- Write a custom tool whose handler runs in the application.
- Disable the capability at configuration or tool-registration level.
- Configure a supported remote HTTP transport with authentication and deployment controls.
- Return a concise task-relevant projection from the tool handler.
Explanation
Correct Answer
D:To reduce context bloat caused by a tool returning a thousand irrelevant fields in this function-calling service, tool output is subsequently consumed as model context; minimizing it reduces both token cost and distractions; the proposed implementation directly supplies that capability.
Incorrect Answers
A:The proposed alternative helps to invoke a proprietary internal pricing database unavailable through a built-in feature because custom tools expose application-specific actions while retaining control over authentication, validation, and network access; that mechanism does not meet the different requirement to reduce context bloat caused by a tool returning a thousand irrelevant fields.
B:The proposed alternative helps to remove an agent's access to an obsolete integration without rewriting all its task prompts because actual registered capabilities define what may execute; changing descriptive prompt text alone does not revoke a previously available tool; that mechanism does not meet the different requirement to reduce context bloat caused by a tool returning a thousand irrelevant fields.
C:The proposed alternative helps to expose an MCP server to clients across a network boundary because remote connectivity needs network routing, identity, and authorization; a local stdio subprocess is not a remotely reachable service; that mechanism does not meet the different requirement to reduce context bloat caused by a tool returning a thousand irrelevant fields.
Question 570
A design review for a high-volume Messages API gateway asks for support to avoid silently treating a broken streaming connection as a complete answer, together with a way to use provider-hosted search or execution without implementing the tool body locally. Select the two appropriate changes. Choose TWO.
- Parse response content by block type and follow current thinking-field rules.
- Handle stop_reason max_tokens as truncation and continue or revise the limit.
- Configure an available server-side tool with the required API definition.
- Detect transport interruption and mark incomplete generated content.
- Enable streaming and consume server-sent events in order.
Explanation
Correct Answers
C:To use provider-hosted search or execution without implementing the tool body locally as part of a design that must also avoid silently treating a broken streaming connection as a complete answer, provider-executed tools are handled in Anthropic's environment and differ from client functions that the caller must run; this is why the action is one of the two required selections.
D:To avoid silently treating a broken streaming connection as a complete answer as part of a design that must also use provider-hosted search or execution without implementing the tool body locally, network termination before a final message event can leave a partial response, which needs explicit recovery rather than success handling; this is why the action is one of the two required selections.
Incorrect Answers
A:This alternative serves the separate goal to handle a response containing reasoning-related blocks without confusing them with ordinary final text because structured content may contain distinct block types and signatures; a parser that assumes only plain text can discard important protocol data; the required outcomes are instead to avoid silently treating a broken streaming connection as a complete answer and to use provider-hosted search or execution without implementing the tool body locally.
B:This alternative serves the separate goal to recover when a response ends because its generation budget was reached because a truncated completion may contain unfinished content, so treating it as a validated final answer risks losing required fields; the required outcomes are instead to avoid silently treating a broken streaming connection as a complete answer and to use provider-hosted search or execution without implementing the tool body locally.
E:This alternative serves the separate goal to show generated text progressively instead of waiting for the full response because streaming transmits partial deltas and completion events, allowing earlier display while requiring the client to assemble content safely; the required outcomes are instead to avoid silently treating a broken streaming connection as a complete answer and to use provider-hosted search or execution without implementing the tool body locally.