Question 700
In a developer productivity assistant, the design must ensure that dependent analysis begins only when source retrieval is complete and that a truncated response is not mistaken for successful completion. Choose TWO independent changes that directly meet these two requirements.
- handle max_tokens as truncation rather than end_turn
- run retrieval before analysis and pass its validated artifacts
- start the analysis and retrieval agents together without shared results
- interpret every non-tool stop reason as a successful final answer
Explanation
Correct Answers
A:For a developer productivity assistant, reaching an output limit is a stop condition but does not establish task completion, which directly establishes that a truncated response is not mistaken for successful completion; the separate requirement that dependent analysis begins only when source retrieval is complete is handled by a different action.
B:For a developer productivity assistant, a downstream agent needs actual upstream findings rather than an assumed future result, which directly establishes that dependent analysis begins only when source retrieval is complete; the separate requirement that a truncated response is not mistaken for successful completion is handled by a different action.
Incorrect Answers
C:For a developer productivity assistant, the analysis branch cannot evaluate evidence it has not received, so this option cannot establish that dependent analysis begins only when source retrieval is complete; correctly addressing a truncated response is not mistaken for successful completion elsewhere does not repair that deficiency.
D:For a developer productivity assistant, a length-limited result may be incomplete or unusable, so this option cannot establish that a truncated response is not mistaken for successful completion; correctly addressing dependent analysis begins only when source retrieval is complete elsewhere does not repair that deficiency.
Question 701
When designing a build verification pipeline, the requirement is not satisfied unless a coding agent inspects a known source file without changing it and only one approved write tool is callable in a sensitive stage. Which implementation addresses the two independent conditions?
- use Read on the identified file; and let the model select from a broad global list of tools
- run arbitrary shell commands to modify the file during inspection; and restrict the stage's allowed tool set and verify permissions in the executor
- use Read on the identified file; and restrict the stage's allowed tool set and verify permissions in the executor
- run arbitrary shell commands to modify the file during inspection; and let the model select from a broad global list of tools
Explanation
Correct Answer
C:In a build verification pipeline, file reading provides exact content for analysis without executing the project; alongside this, capability scoping narrows possible side effects before model selection, satisfying the distinct requirements that a coding agent inspects a known source file without changing it and that only one approved write tool is callable in a sensitive stage.
Incorrect Answers
A:In a build verification pipeline, file reading provides exact content for analysis without executing the project, but excess capabilities create avoidable routing and authorization risk; therefore the separate requirement that only one approved write tool is callable in a sensitive stage would remain unsatisfied.
B:In a build verification pipeline, capability scoping narrows possible side effects before model selection, but mutation is unnecessary for a read-only code review; therefore the separate requirement that a coding agent inspects a known source file without changing it would remain unsatisfied.
D:In a build verification pipeline, mutation is unnecessary for a read-only code review and excess capabilities create avoidable routing and authorization risk; the design would fail both conditions: a coding agent inspects a known source file without changing it and only one approved write tool is callable in a sensitive stage.
Question 702
A team operating a technical due-diligence report generator wants to establish two separate behaviors: a conclusion supported by model agreement is also tested against source evidence; in addition, a parser preserves the specified output structure across cases. Which plan implements both?
- promote agreement alone to a guaranteed correctness signal; and mix contradictory schemas in the example set
- require source and business-rule checks regardless of model agreement; and keep each demonstration's field order and types consistent
- promote agreement alone to a guaranteed correctness signal; and keep each demonstration's field order and types consistent
- require source and business-rule checks regardless of model agreement; and mix contradictory schemas in the example set
Explanation
Correct Answer
B:In a technical due-diligence report generator, correlated model errors can survive consensus without external verification; alongside this, coherent exemplars teach a stable mapping rather than competing formats, satisfying the distinct requirements that a conclusion supported by model agreement is also tested against source evidence and that a parser preserves the specified output structure across cases.
Incorrect Answers
A:In a technical due-diligence report generator, shared biases can produce identical wrong conclusions and inconsistent samples can make expected output ambiguous; the design would fail both conditions: a conclusion supported by model agreement is also tested against source evidence and a parser preserves the specified output structure across cases.
C:In a technical due-diligence report generator, coherent exemplars teach a stable mapping rather than competing formats, but shared biases can produce identical wrong conclusions; therefore the separate requirement that a conclusion supported by model agreement is also tested against source evidence would remain unsatisfied.
D:In a technical due-diligence report generator, correlated model errors can survive consensus without external verification, but inconsistent samples can make expected output ambiguous; therefore the separate requirement that a parser preserves the specified output structure across cases would remain unsatisfied.
Question 703
In a vendor onboarding document pipeline, the design must ensure that temporary service faults can be retried selectively and that the assistant can use external service capabilities via MCP. Choose TWO independent changes that directly meet these two requirements.
- retry every error with the same request indefinitely
- describe the external API in a prompt without connecting any tool
- include a retryable classification and bounded retry advice
- configure a compatible MCP server and expose its declared tools to the client
Explanation
Correct Answers
C:For a vendor onboarding document pipeline, typed transient failures allow backoff without treating permanent errors the same way, which directly establishes that temporary service faults can be retried selectively; the separate requirement that the assistant can use external service capabilities via MCP is handled by a different action.
D:For a vendor onboarding document pipeline, MCP bridges a client to provider-defined capabilities rather than inventing a Claude built-in, which directly establishes that the assistant can use external service capabilities via MCP; the separate requirement that temporary service faults can be retried selectively is handled by a different action.
Incorrect Answers
A:For a vendor onboarding document pipeline, permission and validation failures will not heal through repetition, so this option cannot establish that temporary service faults can be retried selectively; correctly addressing the assistant can use external service capabilities via MCP elsewhere does not repair that deficiency.
B:For a vendor onboarding document pipeline, instructions cannot create an executable integration endpoint, so this option cannot establish that the assistant can use external service capabilities via MCP; correctly addressing temporary service faults can be retried selectively elsewhere does not repair that deficiency.
Question 704
An incident postmortem for a quarterly market-intelligence coordinator requires controls so that a failed retrieval agent does not masquerade as having found no matches and that a formatter handles diverse legitimate examples consistently. Which paired remedy covers both failure modes?
- convert every retrieval failure into an empty list; and give a small set of representative input-output demonstrations
- return a typed failure status with its origin to the coordinator; and give a small set of representative input-output demonstrations
- return a typed failure status with its origin to the coordinator; and supply only a bare instruction to keep formatting consistent
- convert every retrieval failure into an empty list; and supply only a bare instruction to keep formatting consistent
- implement only give a small set of representative input-output demonstrations; make no corresponding change for the first requirement
Explanation
Correct Answer
B:In a quarterly market-intelligence coordinator, an operational error must remain distinguishable from a valid empty result; alongside this, few-shot examples show the intended mapping more concretely than broad style adjectives, satisfying the distinct requirements that a failed retrieval agent does not masquerade as having found no matches and that a formatter handles diverse legitimate examples consistently.
Incorrect Answers
A:In a quarterly market-intelligence coordinator, few-shot examples show the intended mapping more concretely than broad style adjectives, but the final report could incorrectly claim no sources exist; therefore the separate requirement that a failed retrieval agent does not masquerade as having found no matches would remain unsatisfied.
C:In a quarterly market-intelligence coordinator, an operational error must remain distinguishable from a valid empty result, but abstract instructions leave many output choices open; therefore the separate requirement that a formatter handles diverse legitimate examples consistently would remain unsatisfied.
D:In a quarterly market-intelligence coordinator, the final report could incorrectly claim no sources exist and abstract instructions leave many output choices open; the design would fail both conditions: a failed retrieval agent does not masquerade as having found no matches and a formatter handles diverse legitimate examples consistently.
E:In a quarterly market-intelligence coordinator, few-shot examples show the intended mapping more concretely than broad style adjectives, but the independent mechanism for the other criterion is absent: an operational error must remain distinguishable from a valid empty result; as written, the design does not ensure that a failed retrieval agent does not masquerade as having found no matches.
Question 705
The architect for an automated CI review service needs to ensure that a destructive tool receives human authorization at execution time. What should the team implement?
- interrupt or deny the tool until an authorization callback approves it
- use an observation-only hook for telemetry when rewriting is unnecessary
- check resolved target paths in the pre-tool policy gate
- collect a general approval at the start of the conversation
Explanation
Correct Answer
A:For an automated CI review service, execution gates make the approval binding on the selected action and inputs; this directly establishes that a destructive tool receives human authorization at execution time instead of relying on an unrelated workflow control.
Incorrect Answers
B:Although separating instrumentation from mutation reduces the risk of corrupting evidence, that only supports the distinct goal that tool-call logging does not modify a successful tool result, not the requirement that a destructive tool receives human authorization at execution time in an automated CI review service.
C:Although a boundary check on the actual tool argument enforces the permitted filesystem scope, that only supports the distinct goal that file writes are blocked outside an approved workspace, not the requirement that a destructive tool receives human authorization at execution time in an automated CI review service.
D:In an automated CI review service, blanket consent does not verify the later high-risk operation; as a result, the stated requirement that a destructive tool receives human authorization at execution time would still not be met.
Question 706
An audit of a code policy conformance job asks how to make sure that multiple subagent outputs can be traced to their individual tasks. Which implementation should be selected?
- pass the necessary task context and references explicitly at invocation
- define each subagent role and expose the minimum matching tools
- attach task IDs and source references to returned artifacts
- strip origin metadata from each response before merging
Explanation
Correct Answer
C:For a code policy conformance job, traceable results let the coordinator identify which delegation produced each claim; this directly establishes that multiple subagent outputs can be traced to their individual tasks instead of relying on an unrelated workflow control.
Incorrect Answers
A:Although subagents do not automatically inherit the coordinator's entire conversational history, that only supports the distinct goal that a delegated agent receives the background it needs to act, not the requirement that multiple subagent outputs can be traced to their individual tasks in a code policy conformance job.
B:Although tool scoping shrinks both decision confusion and the blast radius of delegated actions, that only supports the distinct goal that specialized agents work with only relevant permissions, not the requirement that multiple subagent outputs can be traced to their individual tasks in a code policy conformance job.
D:In a code policy conformance job, source-less aggregation impairs diagnosis and fact checking; as a result, the stated requirement that multiple subagent outputs can be traced to their individual tasks would still not be met.
Question 707
Before expanding a research PDF metadata pipeline, the owner must demonstrate that the team can diagnose which fields repeatedly fail extraction, while also demonstrating that few-shot material does not contain contradictory answer labels. Which design provides both controls?
- record validation failure codes and field paths per attempt; and keep historical examples even after the labeling policy changes
- log only the final success-or-failure boolean; and audit examples against the current decision rubric before use
- log only the final success-or-failure boolean; and keep historical examples even after the labeling policy changes
- record validation failure codes and field paths per attempt; and audit examples against the current decision rubric before use
Explanation
Correct Answer
D:In a research PDF metadata pipeline, granular observability reveals schema weaknesses and document patterns; alongside this, consistent labels reinforce the intended criteria rather than training ambiguity into the prompt, satisfying the distinct requirements that the team can diagnose which fields repeatedly fail extraction and that few-shot material does not contain contradictory answer labels.
Incorrect Answers
A:In a research PDF metadata pipeline, granular observability reveals schema weaknesses and document patterns, but stale demonstrations can contradict the updated standard; therefore the separate requirement that few-shot material does not contain contradictory answer labels would remain unsatisfied.
B:In a research PDF metadata pipeline, consistent labels reinforce the intended criteria rather than training ambiguity into the prompt, but an aggregate flag hides the source of extraction errors; therefore the separate requirement that the team can diagnose which fields repeatedly fail extraction would remain unsatisfied.
C:In a research PDF metadata pipeline, an aggregate flag hides the source of extraction errors and stale demonstrations can contradict the updated standard; the design would fail both conditions: the team can diagnose which fields repeatedly fail extraction and few-shot material does not contain contradictory answer labels.
Question 708
For a typed API generator project, a reviewer identifies two independent obligations: a delegated agent receives the background it needs to act and the review produces a small, auditable patch. Which proposal should be approved?
- pass the necessary task context and references explicitly at invocation; and iterate with focused diffs rather than wholesale regeneration
- assume the child agent sees all previous coordinator turns; and replace every file in the component to ensure consistency
- pass the necessary task context and references explicitly at invocation; and replace every file in the component to ensure consistency
- assume the child agent sees all previous coordinator turns; and iterate with focused diffs rather than wholesale regeneration
Explanation
Correct Answer
A:In a typed API generator project, subagents do not automatically inherit the coordinator's entire conversational history; alongside this, narrow changes are easier to test and evaluate against explicit requirements, satisfying the distinct requirements that a delegated agent receives the background it needs to act and that the review produces a small, auditable patch.
Incorrect Answers
B:In a typed API generator project, implicit context inheritance makes delegation fragile and incomplete and unnecessary regeneration can erase working behavior; the design would fail both conditions: a delegated agent receives the background it needs to act and the review produces a small, auditable patch.
C:In a typed API generator project, subagents do not automatically inherit the coordinator's entire conversational history, but unnecessary regeneration can erase working behavior; therefore the separate requirement that the review produces a small, auditable patch would remain unsatisfied.
D:In a typed API generator project, narrow changes are easier to test and evaluate against explicit requirements, but implicit context inheritance makes delegation fragile and incomplete; therefore the separate requirement that a delegated agent receives the background it needs to act would remain unsatisfied.
Question 709
Before expanding a policy document normalization service, the owner must demonstrate that a failed high-stakes step prevents a false final success signal, while also demonstrating that persistent ambiguous data receives human review. Which design provides both controls?
- ignore errors after the first successful agent completion; and stop after defined retries and route unresolved cases to a reviewer
- ignore errors after the first successful agent completion; and keep retrying until the model finally emits something that passes
- enforce required-stage completion before publishing a success status; and keep retrying until the model finally emits something that passes
- enforce required-stage completion before publishing a success status; and stop after defined retries and route unresolved cases to a reviewer
Explanation
Correct Answer
D:In a policy document normalization service, a final gate must account for every mandatory stage rather than aggregate optimistic text; alongside this, a bounded escalation path prevents infinite loops and invented resolutions, satisfying the distinct requirements that a failed high-stakes step prevents a false final success signal and that persistent ambiguous data receives human review.
Incorrect Answers
A:In a policy document normalization service, a bounded escalation path prevents infinite loops and invented resolutions, but later failures can invalidate the claimed result; therefore the separate requirement that a failed high-stakes step prevents a false final success signal would remain unsatisfied.
B:In a policy document normalization service, later failures can invalidate the claimed result and unlimited retries waste resources and can reinforce guesses; the design would fail both conditions: a failed high-stakes step prevents a false final success signal and persistent ambiguous data receives human review.
C:In a policy document normalization service, a final gate must account for every mandatory stage rather than aggregate optimistic text, but unlimited retries waste resources and can reinforce guesses; therefore the separate requirement that persistent ambiguous data receives human review would remain unsatisfied.
Question 710
Production readiness for a procurement document processing system depends on ensuring that a function call accepts only valid argument shapes and that conflicting model assessments of a disputed claim are reconciled against evidence. Select the most appropriate combined approach.
- allow arbitrary keys and silently coerce incorrect types; and resolve the dispute against cited source evidence rather than majority vote
- define the tool input schema, enable strict tool use where available, and validate inputs in the handler; and select whichever answer appears in the most generated responses
- allow arbitrary keys and silently coerce incorrect types; and select whichever answer appears in the most generated responses
- define the tool input schema, enable strict tool use where available, and validate inputs in the handler; and resolve the dispute against cited source evidence rather than majority vote
Explanation
Correct Answer
D:In a procurement document processing system, strict tool inputs constrain generation, and handler validation provides a final enforceable argument boundary; alongside this, independent verification distinguishes a well-grounded minority view from shared guessing, satisfying the distinct requirements that a function call accepts only valid argument shapes and that conflicting model assessments of a disputed claim are reconciled against evidence.
Incorrect Answers
A:In a procurement document processing system, independent verification distinguishes a well-grounded minority view from shared guessing, but unvalidated coercion can change intended operations or silently drop important fields; therefore the separate requirement that a function call accepts only valid argument shapes would remain unsatisfied.
B:In a procurement document processing system, strict tool inputs constrain generation, and handler validation provides a final enforceable argument boundary, but agreement among models is not a substitute for factual grounding; therefore the separate requirement that conflicting model assessments of a disputed claim are reconciled against evidence would remain unsatisfied.
C:In a procurement document processing system, unvalidated coercion can change intended operations or silently drop important fields and agreement among models is not a substitute for factual grounding; the design would fail both conditions: a function call accepts only valid argument shapes and conflicting model assessments of a disputed claim are reconciled against evidence.