Question 220
An Azure AI workload passes basic health probes but has this issue: An unconstrained agent begins offering account-deletion actions during troubleshooting conversations. Which fix targets it?
- Specify the support role and allowed actions, and constrain available tools to that role
- Delimit external evidence and apply instruction-hierarchy and injection controls
- Enforce an extraction-to-analysis handoff with schema validation at the boundary
- Define and test compensating operations for successfully completed steps when later steps fail
- Persist transactional state and idempotency keys outside the language-model transcript
Explanation
Correct Answer
A:The fault signal is that an unconstrained agent begins offering account-deletion actions during troubleshooting conversations; the targeted fix works because a clear role and allowed-goal scope guide planning; enforceable permissions remain necessary at tool boundaries, restoring the ability to keep a support agent focused on diagnosing product issues rather than taking unrelated administrative action.
Incorrect Answers
B:If the failing function were to keep untrusted customer-uploaded text from redefining the agent's permissions, this change could be relevant because trusted control instructions must have higher precedence than external content; retrieved pages are data, not agent policy; the actual symptom is that an unconstrained agent begins offering account-deletion actions during troubleshooting conversations, which calls for a different repair.
C:If the failing function were to require one specialist's validated output before another starts its dependent work, this change could be relevant because a sequential handoff makes intermediate contracts explicit; parallel execution would violate a data dependency; the actual symptom is that an unconstrained agent begins offering account-deletion actions during troubleshooting conversations, which calls for a different repair.
D:If the failing function were to recover from a partially completed workflow when a later operation fails, this change could be relevant because a compensating action undoes or offsets previously committed steps where a transaction cannot be atomic across services; the actual symptom is that an unconstrained agent begins offering account-deletion actions during troubleshooting conversations, which calls for a different repair.
E:If the failing function were to resume a long-running workflow safely after an inference retry, this change could be relevant because durable workflow state records completed steps and pending approvals; relying only on the model's last text risks duplicate side effects; the actual symptom is that an unconstrained agent begins offering account-deletion actions during troubleshooting conversations, which calls for a different repair.
Question 221
The engineering change request is to separate developer rules from retrieved instructions. What is the most direct acceptance criterion?
- Replay interrupted workflows and verify each external side effect occurs once
- Test out-of-scope requests for proper refusal or delegated escalation
- Red-team embedded document instructions and inspect unauthorized tool-call attempts
- Compute end-to-end task completion and false-success claim rates
Explanation
Correct Answer
C:The accepted change is to separate developer rules from retrieved instructions; this evaluation checks whether it can keep untrusted customer-uploaded text from redefining the agent's permissions, since trusted control instructions must have higher precedence than external content; retrieved pages are data, not agent policy.
Incorrect Answers
A:This test concerns the ability to resume a long-running workflow safely after an inference retry, for which durable workflow state records completed steps and pending approvals; relying only on the model's last text risks duplicate side effects; it does not establish whether the change to separate developer rules from retrieved instructions works for the current release.
B:This test concerns the ability to keep a support agent focused on diagnosing product issues rather than taking unrelated administrative action, for which a clear role and allowed-goal scope guide planning; enforceable permissions remain necessary at tool boundaries; it does not establish whether the change to separate developer rules from retrieved instructions works for the current release.
D:This test concerns the ability to determine whether the agent actually achieved a multi-step business goal, for which task-success tests compare final system state with expected outcomes; prose fluency does not prove work was completed; it does not establish whether the change to separate developer rules from retrieved instructions works for the current release.
Question 222
A design review proposes to summarize old dialogue for context-window management. Which measurable check would establish success on real inputs?
- Inspect memory contents and verify secrets are absent after session termination
- Test pronoun resolution across turns and verify session boundaries
- Check recall of critical decisions and track context-token savings over long sessions
- Locate the first failing step and measure time spent by workflow stage
Explanation
Correct Answer
C:The accepted change is to summarize old dialogue for context-window management; this evaluation checks whether it can retain key decisions after a long conversation becomes too large to replay verbatim, since a validated rolling summary compresses history while preserving commitments and identifiers; careless compression can discard safety constraints.
Incorrect Answers
A:This test concerns the ability to avoid retaining authentication tokens or temporary PII inside agent memory longer than needed, for which short-lived state and purpose-based retention protect secrets and sensitive data; summarization is not a substitute for removal; it does not establish whether the change to summarize old dialogue for context-window management works for the current release.
B:This test concerns the ability to resolve 'those two records' by retaining the preceding selection context, for which conversation tracking gives referential continuity across turns, but the context should be bounded to the correct user session; it does not establish whether the change to summarize old dialogue for context-window management works for the current release.
D:This test concerns the ability to find where a lengthy agent plan diverges from intended workflow, for which per-step tracing captures model requests, tool parameters, statuses and transitions; a final response log hides the path taken; it does not establish whether the change to summarize old dialogue for context-window management works for the current release.
Question 223
An engineer receives this incident report: The risk-analysis agent starts before the document-extraction agent has produced the validated exposure data. What should be changed?
- Pass sanitized aggregate outputs and enforce each agent's data entitlements
- Enforce an extraction-to-analysis handoff with schema validation at the boundary
- Categorize traces by root-cause stage and track regressions by version
- Classify request intent and dispatch to an authorized specialist with scoped capabilities
Explanation
Correct Answer
B:The fault signal is that the risk-analysis agent starts before the document-extraction agent has produced the validated exposure data; the targeted fix works because a sequential handoff makes intermediate contracts explicit; parallel execution would violate a data dependency, restoring the ability to require one specialist's validated output before another starts its dependent work.
Incorrect Answers
A:If the failing function were to stop one agent from receiving private data that another agent needed only for its subtask, this change could be relevant because scoped messages and redacted handoff contracts minimize sensitive context propagation across multiple agents; the actual symptom is that the risk-analysis agent starts before the document-extraction agent has produced the validated exposure data, which calls for a different repair.
C:If the failing function were to prioritize corrections across retrieval misses, malformed tools, policy blocks, and model planning errors, this change could be relevant because failure classification links observed symptoms to the component that can be improved; a single aggregate error rate obscures causes; the actual symptom is that the risk-analysis agent starts before the document-extraction agent has produced the validated exposure data, which calls for a different repair.
D:If the failing function were to route finance, operations, and security questions to different bounded expert agents, this change could be relevant because a router delegates to an appropriate specialist with a narrow tool set; a single all-powerful agent is harder to govern; the actual symptom is that the risk-analysis agent starts before the document-extraction agent has produced the validated exposure data, which calls for a different repair.
Question 224
A cloud engineering team must enable an agent to read key fields from photographed or scanned paperwork. What is the most appropriate Azure AI implementation?
- Preserve short-term dialogue references
- Add code execution for bounded data calculations
- Use a content-extraction tool for uploaded forms
- Register a read-only API tool for live status
Explanation
Correct Answer
C:For the stated requirement to enable an agent to read key fields from photographed or scanned paperwork, document extraction tools convert visual text and layout into structured data suitable for subsequent agent reasoning; this is the mechanism that the system must implement.
Incorrect Answers
A:For a system whose objective is to resolve 'those two records' by retaining the preceding selection context, conversation tracking gives referential continuity across turns, but the context should be bounded to the correct user session; this does not directly meet the required outcome to enable an agent to read key fields from photographed or scanned paperwork.
B:For a system whose objective is to let an agent aggregate a supplied numeric CSV where exact computation is important, a sandboxed execution tool performs calculations deterministically; executable code needs strict isolation and resource limits; this does not directly meet the required outcome to enable an agent to read key fields from photographed or scanned paperwork.
D:For a system whose objective is to let an agent check a shipment's present location without permission to modify the order, a scoped API tool obtains current transactional facts while maintaining least-privilege read access; this does not directly meet the required outcome to enable an agent to read key fields from photographed or scanned paperwork.
Question 225
A project needs to complete all of these distinct tasks: (1) keep a support agent focused on diagnosing product issues rather than taking unrelated administrative action; (2) keep untrusted customer-uploaded text from redefining the agent's permissions; and (3) resume a long-running workflow safely after an inference retry. Select one suitable capability for each task. Choose THREE.
- Design a typed tool schema for actions
- Define an agent's bounded role and objective
- Maintain task state apart from transient model output
- Separate developer rules from retrieved instructions
- Use a content-extraction tool for uploaded forms
- Limit information passed between specialist agents
Explanation
Correct Answers
B:For the stated requirement to keep a support agent focused on diagnosing product issues rather than taking unrelated administrative action, a clear role and allowed-goal scope guide planning; enforceable permissions remain necessary at tool boundaries; this is the mechanism that the system must implement. This choice satisfies the specific deliverable to keep a support agent focused on diagnosing product issues rather than taking unrelated administrative action, while the other deliverables require separate capabilities.
C:For the stated requirement to resume a long-running workflow safely after an inference retry, durable workflow state records completed steps and pending approvals; relying only on the model's last text risks duplicate side effects; this is the mechanism that the system must implement. This choice satisfies the specific deliverable to resume a long-running workflow safely after an inference retry, while the other deliverables require separate capabilities.
D:For the stated requirement to keep untrusted customer-uploaded text from redefining the agent's permissions, trusted control instructions must have higher precedence than external content; retrieved pages are data, not agent policy; this is the mechanism that the system must implement. This choice satisfies the specific deliverable to keep untrusted customer-uploaded text from redefining the agent's permissions, while the other deliverables require separate capabilities.
Incorrect Answers
A:For a system whose objective is to make the parameters of an account-lookup function explicit to the model and executor, a typed tool contract constrains argument names and shapes and simplifies validation; authorization still occurs on execution; this does not directly meet the required outcome to keep untrusted customer-uploaded text from redefining the agent's permissions.
E:For a system whose objective is to enable an agent to read key fields from photographed or scanned paperwork, document extraction tools convert visual text and layout into structured data suitable for subsequent agent reasoning; this does not directly meet the required outcome to keep untrusted customer-uploaded text from redefining the agent's permissions.
F:For a system whose objective is to stop one agent from receiving private data that another agent needed only for its subtask, scoped messages and redacted handoff contracts minimize sensitive context propagation across multiple agents; this does not directly meet the required outcome to keep untrusted customer-uploaded text from redefining the agent's permissions.
Question 226
In an architecture review, the key requirement is to let an agent check a shipment's present location without permission to modify the order. Which solution directly satisfies it?
- Orchestrate agent knowledge lookup during a conversation
- Register a read-only API tool for live status
- Separate developer rules from retrieved instructions
- Use a content-extraction tool for uploaded forms
Explanation
Correct Answer
B:For the stated requirement to let an agent check a shipment's present location without permission to modify the order, a scoped API tool obtains current transactional facts while maintaining least-privilege read access; this is the mechanism that the system must implement.
Incorrect Answers
A:For a system whose objective is to invoke governed document search as a tool at the point the agent needs evidence for an answer, an agent run needs explicit retrieval calls and returned context before composing a grounded response; model parameters alone do not fetch current documentation; this does not directly meet the required outcome to let an agent check a shipment's present location without permission to modify the order.
C:For a system whose objective is to keep untrusted customer-uploaded text from redefining the agent's permissions, trusted control instructions must have higher precedence than external content; retrieved pages are data, not agent policy; this does not directly meet the required outcome to let an agent check a shipment's present location without permission to modify the order.
D:For a system whose objective is to enable an agent to read key fields from photographed or scanned paperwork, document extraction tools convert visual text and layout into structured data suitable for subsequent agent reasoning; this does not directly meet the required outcome to let an agent check a shipment's present location without permission to modify the order.
Question 227
Engineers have implemented a change to trace step-level agent decisions and failures. What validation would expose a remaining defect?
- Inject mid-workflow failures and audit the final reconciled business state
- Locate the first failing step and measure time spent by workflow stage
- Compute end-to-end task completion and false-success claim rates
- Use a labeled error taxonomy and compare the distribution before and after each change
Explanation
Correct Answer
B:The accepted change is to trace step-level agent decisions and failures; this evaluation checks whether it can find where a lengthy agent plan diverges from intended workflow, since per-step tracing captures model requests, tool parameters, statuses and transitions; a final response log hides the path taken.
Incorrect Answers
A:This test concerns the ability to recover from a partially completed workflow when a later operation fails, for which a compensating action undoes or offsets previously committed steps where a transaction cannot be atomic across services; it does not establish whether the change to trace step-level agent decisions and failures works for the current release.
C:This test concerns the ability to determine whether the agent actually achieved a multi-step business goal, for which task-success tests compare final system state with expected outcomes; prose fluency does not prove work was completed; it does not establish whether the change to trace step-level agent decisions and failures works for the current release.
D:This test concerns the ability to prioritize corrections across retrieval misses, malformed tools, policy blocks, and model planning errors, for which failure classification links observed symptoms to the component that can be improved; a single aggregate error rate obscures causes; it does not establish whether the change to trace step-level agent decisions and failures works for the current release.
Question 228
The support ticket describes this behavior: A retried agent repeats a payment step because it forgot that the payment call succeeded. Which engineering change should come first?
- Persist transactional state and idempotency keys outside the language-model transcript
- Delimit external evidence and apply instruction-hierarchy and injection controls
- Specify the support role and allowed actions, and constrain available tools to that role
- Enable correlated tracing across decision and tool spans with appropriate redaction
Explanation
Correct Answer
A:The fault signal is that a retried agent repeats a payment step because it forgot that the payment call succeeded; the targeted fix works because durable workflow state records completed steps and pending approvals; relying only on the model's last text risks duplicate side effects, restoring the ability to resume a long-running workflow safely after an inference retry.
Incorrect Answers
B:If the failing function were to keep untrusted customer-uploaded text from redefining the agent's permissions, this change could be relevant because trusted control instructions must have higher precedence than external content; retrieved pages are data, not agent policy; the actual symptom is that a retried agent repeats a payment step because it forgot that the payment call succeeded, which calls for a different repair.
C:If the failing function were to keep a support agent focused on diagnosing product issues rather than taking unrelated administrative action, this change could be relevant because a clear role and allowed-goal scope guide planning; enforceable permissions remain necessary at tool boundaries; the actual symptom is that a retried agent repeats a payment step because it forgot that the payment call succeeded, which calls for a different repair.
D:If the failing function were to find where a lengthy agent plan diverges from intended workflow, this change could be relevant because per-step tracing captures model requests, tool parameters, statuses and transitions; a final response log hides the path taken; the actual symptom is that a retried agent repeats a payment step because it forgot that the payment call succeeded, which calls for a different repair.
Question 229
A development group must retain key decisions after a long conversation becomes too large to replay verbatim. What is the most appropriate Azure AI implementation?
- Summarize old dialogue for context-window management
- Preserve short-term dialogue references
- Use a content-extraction tool for uploaded forms
- Retrieve consented durable preferences when required
Explanation
Correct Answer
A:For the stated requirement to retain key decisions after a long conversation becomes too large to replay verbatim, a validated rolling summary compresses history while preserving commitments and identifiers; careless compression can discard safety constraints; this is the mechanism that the system must implement.
Incorrect Answers
B:For a system whose objective is to resolve 'those two records' by retaining the preceding selection context, conversation tracking gives referential continuity across turns, but the context should be bounded to the correct user session; this does not directly meet the required outcome to retain key decisions after a long conversation becomes too large to replay verbatim.
C:For a system whose objective is to enable an agent to read key fields from photographed or scanned paperwork, document extraction tools convert visual text and layout into structured data suitable for subsequent agent reasoning; this does not directly meet the required outcome to retain key decisions after a long conversation becomes too large to replay verbatim.
D:For a system whose objective is to remember a customer's explicit notification preference across separate sessions, persistent preference storage is distinct from a single chat transcript and requires consent, access control, and deletion workflows; this does not directly meet the required outcome to retain key decisions after a long conversation becomes too large to replay verbatim.
Question 230
Before approving the proposed action to use a content-extraction tool for uploaded forms, which measurable acceptance check should be required?
- Test out-of-scope requests for proper refusal or delegated escalation
- Trace agent runs for actual knowledge-tool invocation and check support for each source-grounded answer
- Compare computed totals against independently verified reference calculations
- Measure field-level extraction accuracy on labeled scans and low-quality images
Explanation
Correct Answer
D:The accepted change is to use a content-extraction tool for uploaded forms; this evaluation checks whether it can enable an agent to read key fields from photographed or scanned paperwork, since document extraction tools convert visual text and layout into structured data suitable for subsequent agent reasoning.
Incorrect Answers
A:This test concerns the ability to keep a support agent focused on diagnosing product issues rather than taking unrelated administrative action, for which a clear role and allowed-goal scope guide planning; enforceable permissions remain necessary at tool boundaries; it does not establish whether the change to use a content-extraction tool for uploaded forms works for the current release.
B:This test concerns the ability to invoke governed document search as a tool at the point the agent needs evidence for an answer, for which an agent run needs explicit retrieval calls and returned context before composing a grounded response; model parameters alone do not fetch current documentation; it does not establish whether the change to use a content-extraction tool for uploaded forms works for the current release.
C:This test concerns the ability to let an agent aggregate a supplied numeric CSV where exact computation is important, for which a sandboxed execution tool performs calculations deterministically; executable code needs strict isolation and resource limits; it does not establish whether the change to use a content-extraction tool for uploaded forms works for the current release.