Question 260
A document processing application is approaching release, but the team still needs to let reviewers trace each summary claim to its authoritative document version. What is the most appropriate implementation?
- Add explicit provenance to retrieved evidence
- Separate model reasoning from authoritative calculations
- Return user-friendly error states
- Keep user tasks in the user message
Explanation
Correct Answer
A:Source identifiers, timestamps, and passage boundaries make citations and later audits possible; this behavior directly supports the need to let reviewers trace each summary claim to its authoritative document version. A fluent answer without lineage cannot be independently verified.
Incorrect Answers
B:Deterministic services can enforce exact formulas and rounding before the model explains the result; the supported outcome is to compute discounts through verified business code rather than generated arithmetic alone, whereas this scenario requires the ability to let reviewers trace each summary claim to its authoritative document version; a plausible model narrative is not a reliable financial ledger.
C:This approach serves the separate need to explain service limits or tool outages without leaking stack traces or secrets because structured error mapping presents actionable feedback while retaining detailed diagnostics in protected logs; it does not establish the outcome to let reviewers trace each summary claim to its authoritative document version, and returning a raw exception can expose internal infrastructure details.
D:Explicit message roles distinguish requested work from higher-priority operational framing; despite being useful when a service must separate an end user's request from the developer's stable operating instructions, it is not a solution to let reviewers trace each summary claim to its authoritative document version; merging everything into one unstructured prompt weakens provenance.
Question 261
A design review identifies a missing capability in an agent extensibility project: it must stop a directory lookup function from also gaining unrestricted database write access. Which change meets that goal?
- Validate generated arguments in application code
- Separate read tools from write tools
- Restrict tool network credentials
- Write task-specific tool descriptions
Explanation
Correct Answer
C:Minimal capabilities reduce risk when generated arguments or retrieved text are hostile; this behavior directly supports the need to stop a directory lookup function from also gaining unrestricted database write access. Better prompt wording does not remove excess infrastructure permissions.
Incorrect Answers
A:This approach serves the separate need to prevent malformed or unauthorized input from executing a customer data query because the executor must check types, ranges, identity, and scope independently of the model; it does not establish the outcome to stop a directory lookup function from also gaining unrestricted database write access, and a valid tool schema by itself is not a complete authorization system.
B:Distinct interfaces simplify least-privilege grants and side-effect controls; despite being useful when a service must require stricter approval for actions that change customer data, it is not a solution to stop a directory lookup function from also gaining unrestricted database write access; a generic execute-anything function makes authorization difficult to enforce.
D:Clear capabilities and side effects guide tool selection without granting new authority; this applies to the different goal to let Claude distinguish a read-only invoice lookup from an invoice update, not the requirement to stop a directory lookup function from also gaining unrestricted database write access; vague overlapping descriptions increase erroneous calls.
Question 262
A developer is reviewing alternatives for a credential management feature that must detect harmful instructions at input, tool-call, and output boundaries. Which option is a defensible fit?
- Use layered checks for unsafe requests
- Enforce least-privilege tool permissions
- Require verified identity for privileged actions
- Sandbox generated code execution
- Run adversarial prompt-injection tests
Explanation
Correct Answer
A:Independent controls catch failures that a single prompt-level rule may miss; this behavior directly supports the need to detect harmful instructions at input, tool-call, and output boundaries. A system prompt alone is not a complete enforcement strategy.
Incorrect Answers
B:When an application needs to let a support assistant read shipment status without issuing refunds, restricted scopes and read-only interfaces limit impact if a tool request is manipulated; the current capability must instead detect harmful instructions at input, tool-call, and output boundaries, while role language in the prompt cannot prevent an overprivileged API token from writing.
C:Authentication establishes the caller before authorization policies consider requested permissions; the supported outcome is to stop an unauthenticated user from invoking a dangerous account-management tool, whereas this scenario requires the ability to detect harmful instructions at input, tool-call, and output boundaries; asking the model to remember a user's name is not identity verification.
D:This approach serves the separate need to inspect proposed scripts without granting them production host privileges because resource, filesystem, and network isolation contain failures and malicious effects; it does not establish the outcome to detect harmful instructions at input, tool-call, and output boundaries, and relying on model intent does not constrain a subprocess.
E:Purposeful hostile test cases probe whether the trust boundary holds under manipulated evidence; despite being useful when a service must find vulnerabilities in a retrieval-augmented assistant before it accesses production tools, it is not a solution to detect harmful instructions at input, tool-call, and output boundaries; ordinary happy-path tests rarely exercise attacker instructions.
Question 263
A workflow automation solution has parallel requirements with separate owners. Workstream one must protect a coordinator's window from huge specialist tool outputs; workstream two must manage a complex migration that requires checkpoints after several tool actions. Which combination meets both?
- First component: Use retrieval before generation. Second component: Use a plan-execute-review loop.
- First component: Use an independent verifier step. Second component: Use retrieval before generation.
- First component: Delegate narrow subtasks with a separate context. Second component: Use a plan-execute-review loop.
- First component: Delegate narrow subtasks with a separate context. Second component: Use an independent verifier step.
Explanation
Correct Answer
C:The first proposal fits its assigned role because each subagent can work in a bounded context and return a concise, verified artifact, while the second proposal fits its assigned role because separating plan, execution, and review makes deviations visible before a final outcome is accepted; their respective acceptance criteria are to protect a coordinator's window from huge specialist tool outputs and to manage a complex migration that requires checkpoints after several tool actions.
Incorrect Answers
A:The first proposal serves the different purpose to answer questions from changing internal documentation with traceable source excerpts instead of the assigned first role because retrieval grounds a model answer in current authoritative text rather than hoping parametric knowledge is current, while the second proposal fits its assigned role because separating plan, execution, and review makes deviations visible before a final outcome is accepted; their respective acceptance criteria are to protect a coordinator's window from huge specialist tool outputs and to manage a complex migration that requires checkpoints after several tool actions.
B:The first proposal serves the different purpose to detect mistakes in an agent-generated SQL migration before running it instead of the assigned first role because a separate validation stage checks the artifact against explicit tests and constraints, limiting self-confirming errors, while the second proposal serves the different purpose to answer questions from changing internal documentation with traceable source excerpts instead of the assigned second role because retrieval grounds a model answer in current authoritative text rather than hoping parametric knowledge is current; their respective acceptance criteria are to protect a coordinator's window from huge specialist tool outputs and to manage a complex migration that requires checkpoints after several tool actions.
D:The first proposal fits its assigned role because each subagent can work in a bounded context and return a concise, verified artifact, while the second proposal serves the different purpose to detect mistakes in an agent-generated SQL migration before running it instead of the assigned second role because a separate validation stage checks the artifact against explicit tests and constraints, limiting self-confirming errors; their respective acceptance criteria are to protect a coordinator's window from huge specialist tool outputs and to manage a complex migration that requires checkpoints after several tool actions.
Question 264
An operations orchestration build contains an upstream function and a separate downstream function. The upstream work must analyze several independent document sections concurrently and combine the results; the downstream work must preserve evidence and data lineage when one agent delegates a subtask to another. Which configuration pair covers both?
- First component: Use parallel fan-out and a reconciliation step. Second component: Use an agentic decision loop.
- First component: Use an agentic decision loop. Second component: Adopt a supervisor with specialist workers.
- First component: Use a deterministic workflow. Second component: Use an agentic decision loop.
- First component: Use parallel fan-out and a reconciliation step. Second component: Pass structured handoff artifacts between workers.
- First component: Adopt a supervisor with specialist workers. Second component: Pass structured handoff artifacts between workers.
Explanation
Correct Answer
D:The first proposal fits its assigned role because parallel tasks reduce critical-path latency when they do not depend on one another; a join step resolves conflicting findings, while the second proposal fits its assigned role because defined input and output schemas make it possible to attribute conclusions and detect missing fields at the handoff; their respective acceptance criteria are to analyze several independent document sections concurrently and combine the results and to preserve evidence and data lineage when one agent delegates a subtask to another.
Incorrect Answers
A:The first proposal fits its assigned role because parallel tasks reduce critical-path latency when they do not depend on one another; a join step resolves conflicting findings, while the second proposal serves the different purpose to investigate an open-ended incident where the next query depends on earlier findings instead of the assigned second role because a model-directed loop can choose subsequent actions based on tool observations until it reaches a stopping condition; their respective acceptance criteria are to analyze several independent document sections concurrently and combine the results and to preserve evidence and data lineage when one agent delegates a subtask to another.
B:The first proposal serves the different purpose to investigate an open-ended incident where the next query depends on earlier findings instead of the assigned first role because a model-directed loop can choose subsequent actions based on tool observations until it reaches a stopping condition, while the second proposal serves the different purpose to coordinate separate billing, compliance, and engineering analyses under one final decision instead of the assigned second role because a managing agent delegates bounded subtasks to specialists and reconciles their outputs against a common objective; their respective acceptance criteria are to analyze several independent document sections concurrently and combine the results and to preserve evidence and data lineage when one agent delegates a subtask to another.
C:The first proposal serves the different purpose to execute an unchanged sequence of extraction, validation, and publishing steps instead of the assigned first role because a fixed pipeline keeps control flow in application code, making execution order reproducible and easy to audit, while the second proposal serves the different purpose to investigate an open-ended incident where the next query depends on earlier findings instead of the assigned second role because a model-directed loop can choose subsequent actions based on tool observations until it reaches a stopping condition; their respective acceptance criteria are to analyze several independent document sections concurrently and combine the results and to preserve evidence and data lineage when one agent delegates a subtask to another.
E:The first proposal serves the different purpose to coordinate separate billing, compliance, and engineering analyses under one final decision instead of the assigned first role because a managing agent delegates bounded subtasks to specialists and reconciles their outputs against a common objective, while the second proposal fits its assigned role because defined input and output schemas make it possible to attribute conclusions and detect missing fields at the handoff; their respective acceptance criteria are to analyze several independent document sections concurrently and combine the results and to preserve evidence and data lineage when one agent delegates a subtask to another.
Question 265
Two features of a document processing system are under review. One feature must debug where a request stalls between the web UI, queue, and Claude gateway; the other must allow older application clients to keep working while a new extraction field is introduced. Which plan matches the respective needs?
- First component: Avoid retrying non-idempotent writes blindly. Second component: Version APIs and response schemas.
- First component: Correlate distributed traces across services. Second component: Version APIs and response schemas.
- First component: Apply dependency injection for external clients. Second component: Avoid retrying non-idempotent writes blindly.
- First component: Correlate distributed traces across services. Second component: Apply dependency injection for external clients.
Explanation
Correct Answer
B:The first proposal fits its assigned role because a trace identifier links timing and errors through multiple process boundaries, while the second proposal fits its assigned role because compatibility-aware contracts and versioned behavior protect consumers from surprise breaking changes; their respective acceptance criteria are to debug where a request stalls between the web UI, queue, and Claude gateway and to allow older application clients to keep working while a new extraction field is introduced.
Incorrect Answers
A:The first proposal serves the different purpose to keep an interrupted invoice-creation request from generating duplicate invoices instead of the assigned first role because retries of side-effecting actions need idempotency keys or transaction reconciliation, while the second proposal fits its assigned role because compatibility-aware contracts and versioned behavior protect consumers from surprise breaking changes; their respective acceptance criteria are to debug where a request stalls between the web UI, queue, and Claude gateway and to allow older application clients to keep working while a new extraction field is introduced.
C:The first proposal serves the different purpose to test application logic without issuing real model calls or payments instead of the assigned first role because injected interfaces allow deterministic stubs and reduce coupling to vendors, while the second proposal serves the different purpose to keep an interrupted invoice-creation request from generating duplicate invoices instead of the assigned second role because retries of side-effecting actions need idempotency keys or transaction reconciliation; their respective acceptance criteria are to debug where a request stalls between the web UI, queue, and Claude gateway and to allow older application clients to keep working while a new extraction field is introduced.
D:The first proposal fits its assigned role because a trace identifier links timing and errors through multiple process boundaries, while the second proposal serves the different purpose to test application logic without issuing real model calls or payments instead of the assigned second role because injected interfaces allow deterministic stubs and reduce coupling to vendors; their respective acceptance criteria are to debug where a request stalls between the web UI, queue, and Claude gateway and to allow older application clients to keep working while a new extraction field is introduced.
Question 266
A workflow automation review requires two independent, explicitly defined outcomes: the first workstream must detect mistakes in an agent-generated SQL migration before running it; a second must branch between remediation and escalation based on the result of a security scan. Which TWO actions directly provide these capabilities? Choose TWO.
- Use a graph with conditional edges
- Use an independent verifier step
- Use event logs for tool actions
- Use retrieval before generation
- Use a graceful fallback path
Explanation
Correct Answers
A:Conditional graph transitions encode known branches while allowing model reasoning within selected nodes; this directly supports task 2 to branch between remediation and escalation based on the result of a security scan, not the separate task or tasks to detect mistakes in an agent-generated SQL migration before running it; a linear chain cannot express result-dependent routing cleanly.
B:A separate validation stage checks the artifact against explicit tests and constraints, limiting self-confirming errors; this directly supports task 1 to detect mistakes in an agent-generated SQL migration before running it, not the separate task or tasks to branch between remediation and escalation based on the result of a security scan; asking the original agent whether it is correct is weaker than executable verification.
Incorrect Answers
C:Structured traces connect tool intent, arguments, results, and subsequent decisions for audit and replay; this implements the different outcome to explain why an agent modified a configuration during a failed automated release rather than the required tasks to detect mistakes in an agent-generated SQL migration before running it and to branch between remediation and escalation based on the result of a security scan, and a final chat summary omits the exact causal sequence, so it is not a direct solution to any of the assigned workstreams.
D:Retrieval grounds a model answer in current authoritative text rather than hoping parametric knowledge is current; this implements the different outcome to answer questions from changing internal documentation with traceable source excerpts rather than the required tasks to detect mistakes in an agent-generated SQL migration before running it and to branch between remediation and escalation based on the result of a security scan, and the retrieved text remains untrusted as instructions and needs relevance checks, so it is not a direct solution to any of the assigned workstreams.
E:A bounded failure branch can return a safe partial result or route to a human rather than loop indefinitely; this implements the different outcome to keep a customer service application operating when an expensive specialist agent cannot complete rather than the required tasks to detect mistakes in an agent-generated SQL migration before running it and to branch between remediation and escalation based on the result of a security scan, and unbounded retries do not ensure eventual success, so it is not a direct solution to any of the assigned workstreams.
Question 267
A technical specification for a production API application assigns two jobs to separate modules: module A must produce a user-facing answer after a successful inventory lookup, and module B must let Claude request a read-only inventory lookup with typed arguments. Which two-part solution is appropriate?
- First component: Send tool_result content back to Claude. Second component: Supply a tool definition with input schema.
- First component: Call the Messages API with explicit model and max_tokens. Second component: Supply a tool definition with input schema.
- First component: Send tool_result content back to Claude. Second component: Include an image content block.
- First component: Include an image content block. Second component: Call the Messages API with explicit model and max_tokens.
Explanation
Correct Answer
A:The first proposal fits its assigned role because a result block tied to the tool call returns the external observation to the model for synthesis, while the second proposal fits its assigned role because a declared tool name, description, and schema constrain the interface offered to the model; their respective acceptance criteria are to produce a user-facing answer after a successful inventory lookup and to let Claude request a read-only inventory lookup with typed arguments.
Incorrect Answers
B:The first proposal serves the different purpose to return an answer to a short user request through a backend service instead of the assigned first role because a message request identifies the target model, a message list, and an output-token ceiling, while the second proposal fits its assigned role because a declared tool name, description, and schema constrain the interface offered to the model; their respective acceptance criteria are to produce a user-facing answer after a successful inventory lookup and to let Claude request a read-only inventory lookup with typed arguments.
C:The first proposal fits its assigned role because a result block tied to the tool call returns the external observation to the model for synthesis, while the second proposal serves the different purpose to analyze a photographed shipping label with the supported multimodal Messages input instead of the assigned second role because image inputs are supplied as typed content with appropriate source data or references; their respective acceptance criteria are to produce a user-facing answer after a successful inventory lookup and to let Claude request a read-only inventory lookup with typed arguments.
D:The first proposal serves the different purpose to analyze a photographed shipping label with the supported multimodal Messages input instead of the assigned first role because image inputs are supplied as typed content with appropriate source data or references, while the second proposal serves the different purpose to return an answer to a short user request through a backend service instead of the assigned second role because a message request identifies the target model, a message list, and an output-token ceiling; their respective acceptance criteria are to produce a user-facing answer after a successful inventory lookup and to let Claude request a read-only inventory lookup with typed arguments.
Question 268
During a backend integration rollout, two different failure paths need remediation. The first component must allow older application clients to keep working while a new extraction field is introduced; separately, the second must reproduce a working SDK installation across staging and production. Which changes should be adopted?
- First component: Version APIs and response schemas. Second component: Pin dependencies with lockfiles.
- First component: Version APIs and response schemas. Second component: Enforce concurrency limits.
- First component: Enforce concurrency limits. Second component: Correlate distributed traces across services.
- First component: Correlate distributed traces across services. Second component: Pin dependencies with lockfiles.
Explanation
Correct Answer
A:The first proposal fits its assigned role because compatibility-aware contracts and versioned behavior protect consumers from surprise breaking changes, while the second proposal fits its assigned role because locked versions and hashes constrain drift between build environments; their respective acceptance criteria are to allow older application clients to keep working while a new extraction field is introduced and to reproduce a working SDK installation across staging and production.
Incorrect Answers
B:The first proposal fits its assigned role because compatibility-aware contracts and versioned behavior protect consumers from surprise breaking changes, while the second proposal serves the different purpose to prevent bursts of document jobs from overwhelming rate limits and memory instead of the assigned second role because a semaphore or bounded queue keeps in-flight work within measured capacity; their respective acceptance criteria are to allow older application clients to keep working while a new extraction field is introduced and to reproduce a working SDK installation across staging and production.
C:The first proposal serves the different purpose to prevent bursts of document jobs from overwhelming rate limits and memory instead of the assigned first role because a semaphore or bounded queue keeps in-flight work within measured capacity, while the second proposal serves the different purpose to debug where a request stalls between the web UI, queue, and Claude gateway instead of the assigned second role because a trace identifier links timing and errors through multiple process boundaries; their respective acceptance criteria are to allow older application clients to keep working while a new extraction field is introduced and to reproduce a working SDK installation across staging and production.
D:The first proposal serves the different purpose to debug where a request stalls between the web UI, queue, and Claude gateway instead of the assigned first role because a trace identifier links timing and errors through multiple process boundaries, while the second proposal fits its assigned role because locked versions and hashes constrain drift between build environments; their respective acceptance criteria are to allow older application clients to keep working while a new extraction field is introduced and to reproduce a working SDK installation across staging and production.
Question 269
A team maintaining a document processing needs to let an operator approve a proposed customer-address change before submission. Which implementation most directly satisfies the requirement?
- Treat retrieval results as untrusted data
- Deduplicate application events
- Limit retained context to authorized history
- Expose a review screen before tool-based writes
- Place durable behavior instructions in the system parameter
Explanation
Correct Answer
D:A confirmation UI converts a model suggestion into a separately authorized human action; this behavior directly supports the need to let an operator approve a proposed customer-address change before submission. A conversational promise to double-check does not constitute approval.
Incorrect Answers
A:Retrieved content should be framed as evidence, never upgraded to new tool or system authority; the supported outcome is to prevent a malicious help article from ordering the model to reveal secrets, whereas this scenario requires the ability to let an operator approve a proposed customer-address change before submission; a trustworthy source URL does not make embedded commands authorized.
B:This approach serves the separate need to avoid starting two extraction jobs for the same uploaded document after repeated browser submissions because an idempotent job key lets the server recognize duplicates at its request boundary; it does not establish the outcome to let an operator approve a proposed customer-address change before submission, and model prompting cannot identify all retries arriving from outside the conversation.
C:Context assembly must apply access checks before adding prior turns or retrieved evidence; despite being useful when a service must prevent a new user from inheriting another user's document excerpts, it is not a solution to let an operator approve a proposed customer-address change before submission; prompt-level instructions to respect privacy cannot repair a cross-tenant retrieval bug.
E:System-level guidance supplies application-wide instruction context, separate from each user's task content; this applies to the different goal to ensure an API assistant consistently uses a defined support role across requests, not the requirement to let an operator approve a proposed customer-address change before submission; a string placed in retrieved documents has no authority to redefine the application policy.
Question 270
A release plan for a retrieval pipeline platform has two acceptance checks: component one must prevent a parser crash when the model returns no ordinary text; component two must preserve auditability when an extractor reformats a date or currency. Which proposal passes both?
- First component: Escape generated content in the UI. Second component: Keep raw and normalized values distinct.
- First component: Validate semantic business invariants. Second component: Escape generated content in the UI.
- First component: Handle refusal and empty content explicitly. Second component: Validate semantic business invariants.
- First component: Handle refusal and empty content explicitly. Second component: Keep raw and normalized values distinct.
Explanation
Correct Answer
D:The first proposal fits its assigned role because response handling must branch on stop reason and content block types before decoding data, while the second proposal fits its assigned role because storing the original evidence and validated normalized value supports later reconciliation; their respective acceptance criteria are to prevent a parser crash when the model returns no ordinary text and to preserve auditability when an extractor reformats a date or currency.
Incorrect Answers
A:The first proposal serves the different purpose to show model-produced HTML snippets without allowing browser script injection instead of the assigned first role because output encoding or sanitization protects display contexts from untrusted markup, while the second proposal fits its assigned role because storing the original evidence and validated normalized value supports later reconciliation; their respective acceptance criteria are to prevent a parser crash when the model returns no ordinary text and to preserve auditability when an extractor reformats a date or currency.
B:The first proposal serves the different purpose to prevent a negative invoice total from reaching a payment ledger instead of the assigned first role because domain rules check value ranges and cross-field relationships beyond syntactic shape, while the second proposal serves the different purpose to show model-produced HTML snippets without allowing browser script injection instead of the assigned second role because output encoding or sanitization protects display contexts from untrusted markup; their respective acceptance criteria are to prevent a parser crash when the model returns no ordinary text and to preserve auditability when an extractor reformats a date or currency.
C:The first proposal fits its assigned role because response handling must branch on stop reason and content block types before decoding data, while the second proposal serves the different purpose to prevent a negative invoice total from reaching a payment ledger instead of the assigned second role because domain rules check value ranges and cross-field relationships beyond syntactic shape; their respective acceptance criteria are to prevent a parser crash when the model returns no ordinary text and to preserve auditability when an extractor reformats a date or currency.