Question 270
Two parallel agents edit the same shared variable and overwrite each other's findings. What architecture change helps?
- Regression-test the router against representative existing and new intents
- Enforce handoff limits and an explicit escalation or terminal condition
- Gate the downstream stage on successful completion of its prerequisite
- Isolate branch state and merge outputs deterministically after completion
Explanation
Correct Answer
D:Independent branch outputs avoid races; a controlled merge preserves all required findings. In the multi-step operations assistant, confirm the fix by retesting the affected behavior after isolating branch state and merge outputs deterministically after completion.
Incorrect Answers
A:That change addresses another failure mode: routing classifiers and rules can introduce cross-category mistakes as new paths are introduced; it does not address this case, for which the relevant intervention is to isolate branch state and merge outputs deterministically after completion.
B:Handoff cycles need bounded control flow to avoid infinite delegation and unnecessary model/tool calls in its own workflow, but that is a different correction from the present requirement to isolate branch state and merge outputs deterministically after completion. The reported issue would remain.
C:Although useful elsewhere in the multi-step operations assistant, explicit completion dependencies keep sequential workflows from consuming unvalidated intermediate results; the present case is different and calls for the team to isolate branch state and merge outputs deterministically after completion.
Question 271
A parallel multi-agent workflow combines results from pricing, compliance, and inventory branches. One branch may be slow. Which TWO orchestration decisions are needed before merging results? (Choose TWO.)
- Provide an explicit unknown-intent route to clarification or human triage
- Checkpoint workflow progress and make side-effecting steps idempotent
- Pass the necessary validated context and authorization scope with the handoff
- Specify timeout and partial-result handling for branches that cannot complete
- Define a join condition that waits for all mandatory branches to finish
Explanation
Correct Answers
D:An explicit failure policy determines whether the workflow waits, escalates, or returns a qualified partial answer when a branch stalls.
E:A fan-out workflow needs an explicit synchronization point so the final decision is not made with incomplete required evidence.
Incorrect Answers
A:Although useful elsewhere in the multi-step operations assistant, uncertain routing should not silently select a high-impact specialist without sufficient evidence; the present case is different and calls for the team to define a join condition that waits for all mandatory branches to finish and specify timeout and partial-result handling for branches that cannot complete.
B:That change addresses another failure mode: recoverable orchestration must track completed actions to avoid duplicate writes on retries; it does not address this case, for which the relevant intervention is to define a join condition that waits for all mandatory branches to finish and specify timeout and partial-result handling for branches that cannot complete.
C:A handoff should transfer task-relevant facts without widening the receiving agent's permissions in its own workflow, but that is a different correction from the present requirement to define a join condition that waits for all mandatory branches to finish and specify timeout and partial-result handling for branches that cannot complete. The reported issue would remain.
Question 272
A new topic is added, but routing accuracy degrades for existing tasks. How should changes be assessed?
- Use versioned test fixtures and controlled external-tool responses
- Enforce handoff limits and an explicit escalation or terminal condition
- Isolate branch state and merge outputs deterministically after completion
- Regression-test the router against representative existing and new intents
Explanation
Correct Answer
D:Routing classifiers and rules can introduce cross-category mistakes as new paths are introduced. In the multi-step operations assistant, confirm the fix by retesting the affected behavior after regression-testing the router against representative existing and new intents.
Incorrect Answers
A:That change addresses another failure mode: stable test dependencies distinguish orchestration regressions from variations in outside systems; it does not address this case, for which the relevant intervention is to regression-test the router against representative existing and new intents.
B:Although useful elsewhere in the multi-step operations assistant, handoff cycles need bounded control flow to avoid infinite delegation and unnecessary model/tool calls; the present case is different and calls for the team to regression-test the router against representative existing and new intents.
C:Independent branch outputs avoid races; a controlled merge preserves all required findings in its own workflow, but that is a different correction from the present requirement to regression-test the router against representative existing and new intents. The reported issue would remain.
Question 273
A reporting assistant must retrieve transactions, validate totals, and then draft a report in that order. Which execution pattern fits?
- Run independent analysis branches in parallel and join their results
- Use a routing or triage step based on the user's intent and allowed capabilities
- Use sequential orchestration with explicit dependencies between each stage
- Capture step-level traces, tool inputs, outputs, and terminal outcomes
Explanation
Correct Answer
C:Downstream work depends on earlier validated outputs, so parallel execution risks using incomplete data. In the multi-step operations assistant, confirm the fix by retesting the affected behavior after using sequential orchestration with explicit dependencies between each stage.
Incorrect Answers
A:Although useful elsewhere in the multi-step operations assistant, independent tasks can execute concurrently before a synchronization step produces the combined report; the present case is different and calls for the team to use sequential orchestration with explicit dependencies between each stage.
B:Triage maps a request to the responsible worker rather than running every specialist indiscriminately in its own workflow, but that is a different correction from the present requirement to use sequential orchestration with explicit dependencies between each stage. The reported issue would remain.
D:That change addresses another failure mode: workflow correctness depends on actions taken during the run, not just the fluent final message; it does not address this case, for which the relevant intervention is to use sequential orchestration with explicit dependencies between each stage.
Question 274
A multi-step operations assistant must identify incorrect delegation loops, unsafe actions, and poor end-to-end task success. Reviewers will accept the implementation when reviewers can link a failed outcome to a specific delegation or tool step. Which approach best fulfills that requirement?
- Use Microsoft Agent Framework sequential workflow steps
- Evaluate multi-agent traces and outcomes
- Use Microsoft Agent Framework parallel workflow branches
- Use Microsoft Agent Framework routing or triage
Explanation
Correct Answer
B:Workflow evaluation combines step-level traces with outcome-based measures to test orchestration and reliability. For this multi-step operations assistant, the intended evidence is that reviewers can link a failed outcome to a specific delegation or tool step. Merely increasing the number of agents does not measure whether they accomplish the assigned task.
Incorrect Answers
A:Sequential orchestration enforces dependencies such as research preceding drafting and review preceding publication; it may be used to pass a validated intermediate result from one specialized agent to the next in a required order in a multi-step operations assistant, but the current requirement is instead to identify incorrect delegation loops, unsafe actions, and poor end-to-end task success. That separate capability does not prove that reviewers can link a failed outcome to a specific delegation or tool step; Running dependent stages in parallel can consume incomplete inputs, so selecting that alternative would not meet the stated acceptance check.
C:Parallel orchestration permits independent tasks to execute together and joins outputs for a final synthesis; it may be used to have independent specialists analyze separate evidence sets concurrently before results are merged in a multi-step operations assistant, but the current requirement is instead to identify incorrect delegation loops, unsafe actions, and poor end-to-end task success. That separate capability does not prove that reviewers can link a failed outcome to a specific delegation or tool step; It is unsuitable when later steps require data not produced until an earlier step completes, so selecting that alternative would not meet the stated acceptance check.
D:Routing uses task interpretation to choose a domain agent instead of sending every request through every specialist; it may be used to direct a request to the appropriate specialist based on the detected task in a multi-step operations assistant, but the current requirement is instead to identify incorrect delegation loops, unsafe actions, and poor end-to-end task success. That separate capability does not prove that reviewers can link a failed outcome to a specific delegation or tool step; A fixed sequential chain is unnecessarily rigid when only one branch should run, so selecting that alternative would not meet the stated acceptance check.
Question 275
The solution brief for a multi-step operations assistant requires the system to pass a validated intermediate result from one specialized agent to the next in a required order. A successful test confirms that the second stage begins only after the first stage's output exists. What should the team use?
- Use Microsoft Agent Framework routing or triage
- Evaluate multi-agent traces and outcomes
- Use Microsoft Agent Framework sequential workflow steps
- Use Microsoft Agent Framework parallel workflow branches
Explanation
Correct Answer
C:Sequential orchestration enforces dependencies such as research preceding drafting and review preceding publication. For this multi-step operations assistant, the intended evidence is that the second stage begins only after the first stage's output exists. Running dependent stages in parallel can consume incomplete inputs.
Incorrect Answers
A:Routing uses task interpretation to choose a domain agent instead of sending every request through every specialist; it may be used to direct a request to the appropriate specialist based on the detected task in a multi-step operations assistant, but the current requirement is instead to pass a validated intermediate result from one specialized agent to the next in a required order. That separate capability does not prove that the second stage begins only after the first stage's output exists; A fixed sequential chain is unnecessarily rigid when only one branch should run, so selecting that alternative would not meet the stated acceptance check.
B:Workflow evaluation combines step-level traces with outcome-based measures to test orchestration and reliability; it may be used to identify incorrect delegation loops, unsafe actions, and poor end-to-end task success in a multi-step operations assistant, but the current requirement is instead to pass a validated intermediate result from one specialized agent to the next in a required order. That separate capability does not prove that the second stage begins only after the first stage's output exists; Merely increasing the number of agents does not measure whether they accomplish the assigned task, so selecting that alternative would not meet the stated acceptance check.
D:Parallel orchestration permits independent tasks to execute together and joins outputs for a final synthesis; it may be used to have independent specialists analyze separate evidence sets concurrently before results are merged in a multi-step operations assistant, but the current requirement is instead to pass a validated intermediate result from one specialized agent to the next in a required order. That separate capability does not prove that the second stage begins only after the first stage's output exists; It is unsuitable when later steps require data not produced until an earlier step completes, so selecting that alternative would not meet the stated acceptance check.
Question 276
A sequential workflow resumes after a failure but repeats an already completed payment posting. What should be added?
- Checkpoint workflow progress and make side-effecting steps idempotent
- Provide an explicit unknown-intent route to clarification or human triage
- Score tool authorization and side effects in addition to final-answer quality
- Wait for required branches or apply an explicit timeout and partial-result policy
Explanation
Correct Answer
A:Recoverable orchestration must track completed actions to avoid duplicate writes on retries. In the multi-step operations assistant, confirm the fix by retesting the affected behavior after checkpointing workflow progress and make side-effecting steps idempotent.
Incorrect Answers
B:Uncertain routing should not silently select a high-impact specialist without sufficient evidence in its own workflow, but that is a different correction from the present requirement to checkpoint workflow progress and make side-effecting steps idempotent. The reported issue would remain.
C:Although useful elsewhere in the multi-step operations assistant, a good final answer cannot erase an unsafe intermediate action; the present case is different and calls for the team to checkpoint workflow progress and make side-effecting steps idempotent.
D:That change addresses another failure mode: fan-out/fan-in behavior needs a defined join condition; arbitrary early merge can omit required evidence; it does not address this case, for which the relevant intervention is to checkpoint workflow progress and make side-effecting steps idempotent.
Question 277
The final agent receives a summary before the validation step has completed. What flow design is missing?
- Gate the downstream stage on successful completion of its prerequisite
- Use versioned test fixtures and controlled external-tool responses
- Isolate branch state and merge outputs deterministically after completion
- Regression-test the router against representative existing and new intents
Explanation
Correct Answer
A:Explicit completion dependencies keep sequential workflows from consuming unvalidated intermediate results. In the multi-step operations assistant, confirm the fix by retesting the affected behavior after gating the downstream stage on successful completion of its prerequisite.
Incorrect Answers
B:That change addresses another failure mode: stable test dependencies distinguish orchestration regressions from variations in outside systems; it does not address this case, for which the relevant intervention is to gate the downstream stage on successful completion of its prerequisite.
C:Although useful elsewhere in the multi-step operations assistant, independent branch outputs avoid races; a controlled merge preserves all required findings; the present case is different and calls for the team to gate the downstream stage on successful completion of its prerequisite.
D:Routing classifiers and rules can introduce cross-category mistakes as new paths are introduced in its own workflow, but that is a different correction from the present requirement to gate the downstream stage on successful completion of its prerequisite. The reported issue would remain.
Question 278
A team responsible for a multi-step operations assistant needs to have independent specialists analyze separate evidence sets concurrently before results are merged; the acceptance check is that independent outputs are combined after both branches finish. Which capability should be selected?
- Use Microsoft Agent Framework routing or triage
- Use Microsoft Agent Framework handoff orchestration
- Use Microsoft Agent Framework sequential workflow steps
- Use Microsoft Agent Framework parallel workflow branches
- Evaluate multi-agent traces and outcomes
Explanation
Correct Answer
D:Parallel orchestration permits independent tasks to execute together and joins outputs for a final synthesis. For this multi-step operations assistant, the intended evidence is that independent outputs are combined after both branches finish. It is unsuitable when later steps require data not produced until an earlier step completes.
Incorrect Answers
A:Routing uses task interpretation to choose a domain agent instead of sending every request through every specialist; it may be used to direct a request to the appropriate specialist based on the detected task in a multi-step operations assistant, but the current requirement is instead to have independent specialists analyze separate evidence sets concurrently before results are merged. That separate capability does not prove that independent outputs are combined after both branches finish; A fixed sequential chain is unnecessarily rigid when only one branch should run, so selecting that alternative would not meet the stated acceptance check.
B:Handoff patterns change which agent owns the next step while preserving needed state and authority boundaries; it may be used to transfer an active task and its relevant context to a specialist that takes over the next interaction in a multi-step operations assistant, but the current requirement is instead to have independent specialists analyze separate evidence sets concurrently before results are merged. That separate capability does not prove that independent outputs are combined after both branches finish; A parallel result merge is not the same as transferring conversational responsibility, so selecting that alternative would not meet the stated acceptance check.
C:Sequential orchestration enforces dependencies such as research preceding drafting and review preceding publication; it may be used to pass a validated intermediate result from one specialized agent to the next in a required order in a multi-step operations assistant, but the current requirement is instead to have independent specialists analyze separate evidence sets concurrently before results are merged. That separate capability does not prove that independent outputs are combined after both branches finish; Running dependent stages in parallel can consume incomplete inputs, so selecting that alternative would not meet the stated acceptance check.
E:Workflow evaluation combines step-level traces with outcome-based measures to test orchestration and reliability; it may be used to identify incorrect delegation loops, unsafe actions, and poor end-to-end task success in a multi-step operations assistant, but the current requirement is instead to have independent specialists analyze separate evidence sets concurrently before results are merged. That separate capability does not prove that independent outputs are combined after both branches finish; Merely increasing the number of agents does not measure whether they accomplish the assigned task, so selecting that alternative would not meet the stated acceptance check.
Question 279
An evaluation labels an agent successful even though it made an unauthorized API call before apologizing. What criterion is missing?
- Score tool authorization and side effects in addition to final-answer quality
- Pass the necessary validated context and authorization scope with the handoff
- Provide an explicit unknown-intent route to clarification or human triage
- Checkpoint workflow progress and make side-effecting steps idempotent
Explanation
Correct Answer
A:A good final answer cannot erase an unsafe intermediate action. In the multi-step operations assistant, confirm the fix by retesting the affected behavior after scoring tool authorization and side effects in addition to final-answer quality.
Incorrect Answers
B:Although useful elsewhere in the multi-step operations assistant, a handoff should transfer task-relevant facts without widening the receiving agent's permissions; the present case is different and calls for the team to score tool authorization and side effects in addition to final-answer quality.
C:Uncertain routing should not silently select a high-impact specialist without sufficient evidence in its own workflow, but that is a different correction from the present requirement to score tool authorization and side effects in addition to final-answer quality. The reported issue would remain.
D:That change addresses another failure mode: recoverable orchestration must track completed actions to avoid duplicate writes on retries; it does not address this case, for which the relevant intervention is to score tool authorization and side effects in addition to final-answer quality.
Question 280
A research agent must independently analyze pricing, support tickets, and customer sentiment before combining findings. What orchestration mode is suitable?
- Use sequential orchestration with explicit dependencies between each stage
- Run independent analysis branches in parallel and join their results
- Capture step-level traces, tool inputs, outputs, and terminal outcomes
- Use a routing or triage step based on the user's intent and allowed capabilities
Explanation
Correct Answer
B:Independent tasks can execute concurrently before a synchronization step produces the combined report. In the multi-step operations assistant, confirm the fix by retesting the affected behavior after running independent analysis branches in parallel and join their results.
Incorrect Answers
A:That change addresses another failure mode: downstream work depends on earlier validated outputs, so parallel execution risks using incomplete data; it does not address this case, for which the relevant intervention is to run independent analysis branches in parallel and join their results.
C:Although useful elsewhere in the multi-step operations assistant, workflow correctness depends on actions taken during the run, not just the fluent final message; the present case is different and calls for the team to run independent analysis branches in parallel and join their results.
D:Triage maps a request to the responsible worker rather than running every specialist indiscriminately in its own workflow, but that is a different correction from the present requirement to run independent analysis branches in parallel and join their results. The reported issue would remain.