Question 160
The deployment pipeline runs unattended repository inspections and must not depend on a human interactive session. Which acceptance check evaluates this requirement rather than a proxy?
- The measured evidence confirms that a client discovers approved tools and receives denied responses for prohibited operations.
- The scenario test demonstrates that duplicate deliveries yield one committed document result with correlated processing logs.
- The verified result is that deployment review confirms region, identity scope and network posture meet the architecture baseline.
- The observed outcome is that the same committed pipeline produces comparable artifacts in repeated isolated runs.
Explanation
Correct Answer
D:CLI-based automation fits an existing controlled build environment better than interactive discovery; this checks the situation in which an operations engineer must automate repository checks in a controlled CI execution environment.
Incorrect Answers
A:MCP can standardize discovery and invocation while the server still enforces its own authorization; this applies when a desktop engineering assistant needs standardized access to several internal services using reusable tool definitions, but here the deployment pipeline runs unattended repository inspections and must not depend on a human interactive session.
B:Asynchronous integration decouples producer availability from inference duration and retry behavior; the alternative helps if a partner sends events to a document processor but cannot wait for synchronous model completion, instead of when the deployment pipeline runs unattended repository inspections and must not depend on a human interactive session.
C:Platform choice should align model access with the organization's operational and compliance controls; it is justified if a firm must use its approved cloud identity, networking and deployment controls for model requests, not the case in which the deployment pipeline runs unattended repository inspections and must not depend on a human interactive session.
Question 161
An autonomous research service hands off a task, but the receiving service cannot correlate completion with the requester. Which measurement is the most relevant release criterion?
- The negative-case evaluation confirms that integration tests exercise timeouts, schema validation and error handling through the existing service.
- A controlled trial finds that interoperability tests reject malformed messages and trace delegation back to the initiating user.
- The measured evidence confirms that duplicate deliveries yield one committed document result with correlated processing logs.
- The scenario test demonstrates that an attempted arbitrary command is denied even when embedded in a seemingly valid model tool call.
Explanation
Correct Answer
B:Cross-agent delegation needs agreement on task state, identity and results beyond generic tool discovery; the verification boundary is the one where two autonomous services need structured task delegation with shared correlation IDs.
Incorrect Answers
A:Direct API integration avoids introducing agent protocols for a predictable request-response task; the other case is when a backend application needs a single well-defined model request during an existing HTTP workflow, not the present case where an autonomous research service hands off a task, but the receiving service cannot correlate completion with the requester.
C:Asynchronous integration decouples producer availability from inference duration and retry behavior; this technique fits where a partner sends events to a document processor but cannot wait for synchronous model completion, and not where an autonomous research service hands off a task, but the receiving service cannot correlate completion with the requester.
D:The integration mechanism does not justify expanding application authority beyond necessary actions; the trigger would be that a team wants to expose every shell command as a tool because MCP makes registration convenient, rather than the risk that an autonomous research service hands off a task, but the receiving service cannot correlate completion with the requester.
Question 162
A regulated deployment passes functional tests but bypasses the organization's identity and private-network controls. Which production-like test finding is most convincing?
- The recorded test establishes that deployment review confirms region, identity scope and network posture meet the architecture baseline.
- The representative case test finds that an attempted arbitrary command is denied even when embedded in a seemingly valid model tool call.
- The staged rollout demonstrates that integration tests exercise timeouts, schema validation and error handling through the existing service.
- A controlled trial finds that a client discovers approved tools and receives denied responses for prohibited operations.
- The readiness assessment shows that duplicate deliveries yield one committed document result with correlated processing logs.
Explanation
Correct Answer
A:Platform choice should align model access with the organization's operational and compliance controls; the reviewer is concerned with situations where a firm must use its approved cloud identity, networking and deployment controls for model requests.
Incorrect Answers
B:The integration mechanism does not justify expanding application authority beyond necessary actions; the other workload is one where a team wants to expose every shell command as a tool because MCP makes registration convenient, whereas the release criterion addresses a regulated deployment passes functional tests but bypasses the organization's identity and private-network controls.
C:Direct API integration avoids introducing agent protocols for a predictable request-response task; this applies when a backend application needs a single well-defined model request during an existing HTTP workflow, while the required safeguard concerns a regulated deployment passes functional tests but bypasses the organization's identity and private-network controls.
D:MCP can standardize discovery and invocation while the server still enforces its own authorization; the alternative helps if a desktop engineering assistant needs standardized access to several internal services using reusable tool definitions, not the risk where a regulated deployment passes functional tests but bypasses the organization's identity and private-network controls.
E:Asynchronous integration decouples producer availability from inference duration and retry behavior; it is justified if a partner sends events to a document processor but cannot wait for synchronous model completion, not this situation where a regulated deployment passes functional tests but bypasses the organization's identity and private-network controls.
Question 163
An employee can log into the assistant but retrieve confidential files outside their department. What should be prioritized during architecture review?
- For the primary workstream, propagate signed actor context and authorize each operation at the receiving server.
- For this workflow, apply response-level data minimization and field filtering before context injection.
- For the affected service, enforce source-system authorization on each retrieval using the end user's identity.
- In the project scope, implement short-lived credentials, revocation and access lifecycle controls for delegated sessions.
Explanation
Correct Answer
C:Authentication proves who is calling; per-document authorization determines which evidence may be accessed; the intended outcome is that a user from another department cannot obtain restricted snippets through retrieval or citations.
Incorrect Answers
A:Tool descriptions and model intentions are not an authorization system for external side effects; it is relevant if an MCP connector accepts tool calls without checking which human authorized them, instead of when an employee can log into the assistant but retrieve confidential files outside their department.
B:Authorized retrieval can still overshare; filtering reduces unnecessary sensitive model context; it serves conditions where a tool response includes confidential fields not required by the requested task, not the case in which an employee can log into the assistant but retrieve confidential files outside their department.
D:Long-lived delegated access outlasts valid employment and creates stale privilege exposure; this makes sense when a background job retains access after an employee leaves the company, but this decision concerns an employee can log into the assistant but retrieve confidential files outside their department.
Question 164
A service stores a broad administrator token so a tool can read order statuses. Which response follows from the actual operational constraint?
- Within the delivery plan, replace the token with narrow scoped service credentials and protected secret rotation.
- At the system boundary, apply response-level data minimization and field filtering before context injection.
- For the primary workstream, implement short-lived credentials, revocation and access lifecycle controls for delegated sessions.
- For this workflow, propagate signed actor context and authorize each operation at the receiving server.
Explanation
Correct Answer
A:Least-privilege credentials limit compromise impact and avoid unnecessary administrative capability; the practical acceptance condition is that the tool can read allowed orders but cannot change payments or account roles.
Incorrect Answers
B:Authorized retrieval can still overshare; filtering reduces unnecessary sensitive model context; it serves a setting where a tool response includes confidential fields not required by the requested task, and not where a service stores a broad administrator token so a tool can read order statuses.
C:Long-lived delegated access outlasts valid employment and creates stale privilege exposure; that is defensible if a background job retains access after an employee leaves the company, rather than the risk that a service stores a broad administrator token so a tool can read order statuses.
D:Tool descriptions and model intentions are not an authorization system for external side effects; the other workload is one where an MCP connector accepts tool calls without checking which human authorized them, rather than the condition that a service stores a broad administrator token so a tool can read order statuses.
Question 165
A summarization tool trusts client-provided tenant IDs without checking the authenticated session. Which change targets the constraint rather than a proxy metric?
- For this production path, move enforcement to backend access controls, redaction and secret management.
- For the intended users, bind tenant scope server-side to verified identity claims before issuing any data query.
- Within the delivery plan, implement short-lived credentials, revocation and access lifecycle controls for delegated sessions.
- At the system boundary, propagate signed actor context and authorize each operation at the receiving server.
Explanation
Correct Answer
B:Untrusted request parameters cannot establish data-access authority across customers; the solution is testable when changing the submitted tenant ID does not reveal another organization's records.
Incorrect Answers
A:Prompts guide behavior but do not guarantee prevention of unauthorized disclosure or tool access; the approach helps when a product team treats a prompt instruction to 'never disclose secrets' as the sole data boundary, whereas the release criterion addresses a summarization tool trusts client-provided tenant IDs without checking the authenticated session.
C:Long-lived delegated access outlasts valid employment and creates stale privilege exposure; it is relevant if a background job retains access after an employee leaves the company, not the risk where a summarization tool trusts client-provided tenant IDs without checking the authenticated session.
D:Tool descriptions and model intentions are not an authorization system for external side effects; it serves conditions where an MCP connector accepts tool calls without checking which human authorized them, not this situation where a summarization tool trusts client-provided tenant IDs without checking the authenticated session.
Question 166
An MCP connector accepts tool calls without checking which human authorized them. Which action is justified by the scenario?
- In the proposed design, enforce source-system authorization on each retrieval using the end user's identity.
- Under the reviewed approach, replace the token with narrow scoped service credentials and protected secret rotation.
- For this production path, apply response-level data minimization and field filtering before context injection.
- For the intended users, propagate signed actor context and authorize each operation at the receiving server.
- At this decision point, implement short-lived credentials, revocation and access lifecycle controls for delegated sessions.
Explanation
Correct Answer
D:Tool descriptions and model intentions are not an authorization system for external side effects; a representative trial should prove that calls with missing or insufficient actor claims fail before backend execution.
Incorrect Answers
A:Authentication proves who is calling; per-document authorization determines which evidence may be accessed; that fits cases where an employee can log into the assistant but retrieve confidential files outside their department, not the case in which an MCP connector accepts tool calls without checking which human authorized them.
B:Least-privilege credentials limit compromise impact and avoid unnecessary administrative capability; this control is needed when a service stores a broad administrator token so a tool can read order statuses, whereas the problem presented is that an MCP connector accepts tool calls without checking which human authorized them.
C:Authorized retrieval can still overshare; filtering reduces unnecessary sensitive model context; that would be useful if a tool response includes confidential fields not required by the requested task, but this decision concerns an MCP connector accepts tool calls without checking which human authorized them.
E:Long-lived delegated access outlasts valid employment and creates stale privilege exposure; that is defensible if a background job retains access after an employee leaves the company, but the requested boundary concerns an MCP connector accepts tool calls without checking which human authorized them.
Question 167
A background job retains access after an employee leaves the company. What change directly resolves this production-design issue?
- As a candidate control, replace the token with narrow scoped service credentials and protected secret rotation.
- In the initial solution, implement short-lived credentials, revocation and access lifecycle controls for delegated sessions.
- In the proposed design, apply response-level data minimization and field filtering before context injection.
- Under the reviewed approach, enforce source-system authorization on each retrieval using the end user's identity.
Explanation
Correct Answer
B:Long-lived delegated access outlasts valid employment and creates stale privilege exposure; its effect can be measured when a revoked user token cannot initiate fresh jobs or reuse a session to access protected data.
Incorrect Answers
A:Least-privilege credentials limit compromise impact and avoid unnecessary administrative capability; its scope includes cases where a service stores a broad administrator token so a tool can read order statuses, and not where a background job retains access after an employee leaves the company.
C:Authorized retrieval can still overshare; filtering reduces unnecessary sensitive model context; the approach helps when a tool response includes confidential fields not required by the requested task, rather than the condition that a background job retains access after an employee leaves the company.
D:Authentication proves who is calling; per-document authorization determines which evidence may be accessed; that belongs to situations where an employee can log into the assistant but retrieve confidential files outside their department, while the immediate question involves a background job retains access after an employee leaves the company.
Question 168
A tool response includes confidential fields not required by the requested task. Which proposed design action directly addresses the risk?
- In the proposed implementation, apply response-level data minimization and field filtering before context injection.
- At this decision point, replace the token with narrow scoped service credentials and protected secret rotation.
- As a candidate control, propagate signed actor context and authorize each operation at the receiving server.
- In the initial solution, bind tenant scope server-side to verified identity claims before issuing any data query.
Explanation
Correct Answer
A:Authorized retrieval can still overshare; filtering reduces unnecessary sensitive model context; the solution is testable when response inspections show only approved fields reach model requests and audit traces.
Incorrect Answers
B:Least-privilege credentials limit compromise impact and avoid unnecessary administrative capability; the design helps when a service stores a broad administrator token so a tool can read order statuses, not the risk where a tool response includes confidential fields not required by the requested task.
C:Tool descriptions and model intentions are not an authorization system for external side effects; that fits cases where an MCP connector accepts tool calls without checking which human authorized them, not this situation where a tool response includes confidential fields not required by the requested task.
D:Untrusted request parameters cannot establish data-access authority across customers; this control is needed when a summarization tool trusts client-provided tenant IDs without checking the authenticated session, whereas this scenario involves a tool response includes confidential fields not required by the requested task.
Question 169
A product team treats a prompt instruction to 'never disclose secrets' as the sole data boundary. Which option supplies the needed architectural boundary?
- In the release design, move enforcement to backend access controls, redaction and secret management.
- During architecture review, apply response-level data minimization and field filtering before context injection.
- In the proposed implementation, propagate signed actor context and authorize each operation at the receiving server.
- At this decision point, enforce source-system authorization on each retrieval using the end user's identity.
Explanation
Correct Answer
A:Prompts guide behavior but do not guarantee prevention of unauthorized disclosure or tool access; the intended outcome is that prompt-injection tests cannot retrieve stored secrets because the backend denies the requests.
Incorrect Answers
B:Authorized retrieval can still overshare; filtering reduces unnecessary sensitive model context; it is justified if a tool response includes confidential fields not required by the requested task, but this decision concerns a product team treats a prompt instruction to 'never disclose secrets' as the sole data boundary.
C:Tool descriptions and model intentions are not an authorization system for external side effects; its scope includes cases where an MCP connector accepts tool calls without checking which human authorized them, but the actual risk is that a product team treats a prompt instruction to 'never disclose secrets' as the sole data boundary.
D:Authentication proves who is calling; per-document authorization determines which evidence may be accessed; it targets cases where an employee can log into the assistant but retrieve confidential files outside their department, but the requested boundary concerns a product team treats a prompt instruction to 'never disclose secrets' as the sole data boundary.
Question 170
An authenticated employee manipulates a document identifier and retrieves another department's confidential brief. What independently verifiable observation is needed?
- A production-like test shows that a user from another department cannot obtain restricted snippets through retrieval or citations.
- The release check confirms that response inspections show only approved fields reach model requests and audit traces.
- The reviewer observes that calls with missing or insufficient actor claims fail before backend execution.
- The release report verifies that a revoked user token cannot initiate fresh jobs or reuse a session to access protected data.
Explanation
Correct Answer
A:Authentication proves who is calling; per-document authorization determines which evidence may be accessed; the check directly targets the condition that an employee can log into the assistant but retrieve confidential files outside their department.
Incorrect Answers
B:Authorized retrieval can still overshare; filtering reduces unnecessary sensitive model context; this technique fits where a tool response includes confidential fields not required by the requested task, rather than the condition that an authenticated employee manipulates a document identifier and retrieves another department's confidential brief.
C:Tool descriptions and model intentions are not an authorization system for external side effects; the trigger would be that an MCP connector accepts tool calls without checking which human authorized them, while the immediate question involves an authenticated employee manipulates a document identifier and retrieves another department's confidential brief.
D:Long-lived delegated access outlasts valid employment and creates stale privilege exposure; the design helps when a background job retains access after an employee leaves the company, but the present constraint is that an authenticated employee manipulates a document identifier and retrieves another department's confidential brief.