Question 120 A model-generated answer needs to include a hyperlink to the exact policy source. Which editorial rule applies? Link only to an unrelated marketing landing page Create a realistic-looking but nonexistent URL Use the actual authorized retrieval source and verify the linked document Omit source access checks for all users Explanation Correct Answer C:A factual citation should resolve to the evidence used. Incorrect Answers A:The landing page may not support the stated policy. B:A fabricated link misleads users about provenance. D:The link could expose unauthorized content.
Question 121 A topic should share a locale variable with multiple helper topics during one conversation. Which scope arrangement is appropriate? Create isolated variables with conflicting defaults everywhere Create the locale variable separately with conflicting defaults in every helper topic Use an appropriately scoped global variable or explicit pass-through inputs Write the locale permanently into a public URL Explanation Correct Answer C:Shared state avoids repeated locale prompts across related topics. Incorrect Answers A:Local copies can diverge during one session. B:Independent defaults can select inconsistent languages across a conversation. D:Public URLs are not necessarily the correct state store.
Question 122 A topic uses a Power Fx expression to calculate whether a ticket is overdue. What input validation is required? Ignore null due dates and mark them overdue Use the ticket description length as the date Ensure due-date and current-time values are valid dates in a defined timezone Compare the due date to a literal word like 'tomorrow' Explanation Correct Answer C:Date arithmetic needs proper types and timezone semantics. Incorrect Answers A:Missing data needs an explicit fallback. B:Text length does not measure time. D:Unparsed relative strings are ambiguous.
Question 123 A topic calls a generative answers node and receives no grounded result. How should it proceed? Call a destructive cleanup tool to fix search Return an invented citation to a nonexistent document Repeat the original answer with stronger confidence Offer an uncertainty response and an escalation or targeted follow-up Explanation Correct Answer D:No retrieved evidence means the agent should not claim certainty. Incorrect Answers A:Search misses do not justify data deletion. B:A made-up source is deceptive. C:Confidence does not create missing evidence.
Question 124 A booking agent's topic has an optional phone number field. How should it handle omission? Send the word 'required' as a fabricated phone Drop all required fields when one optional is missing Model it as optional and invoke the API with only contract-appropriate fields Substitute a real customer's phone from a prior session Explanation Correct Answer C:Omission must not be converted into fake phone data. Incorrect Answers A:A placeholder does not satisfy a phone-number schema. B:Optional data should not block otherwise valid requests. D:That risks cross-user data leakage.
Question 125 An agent should display large dynamic data in a readable way rather than one enormous paragraph. Which response strategy helps? Collapse every error and warning into invisible text Output an unfiltered raw API dump to every user Present structured summaries with concise fields and a way to inspect authorized details Replace numeric results with approximate guesses Explanation Correct Answer C:Readable formatting supports comprehension without hiding essential context. Incorrect Answers A:Important exceptions must remain visible. B:Raw payloads can be overwhelming and expose private fields. D:Summarization should preserve actual values.
Question 126 A sales topic requires a region choice, then a conditional region-specific flow. Which sequence fits? Infer region from unrelated browsing history Collect the region in a question node, validate it, and branch to the proper flow Randomly route requests to reduce latency Run every regional write flow simultaneously Explanation Correct Answer B:Explicit input controls and conditional routing connect user choice to the correct backend. Incorrect Answers A:Unverified context may not reflect the user's intended account. C:A routing lottery violates jurisdictional boundaries. D:That may change records in the wrong jurisdictions.
Question 127 A topic must include a correction option when a user notices an incorrect customer ID. Which design is appropriate? Ask for a whole new privileged login without reason Execute the tool first and prohibit changes Silently ignore the corrected ID Allow revising the captured variable before executing the final tool Explanation Correct Answer D:Validated corrections can prevent wrong-account operations. Incorrect Answers A:An unrelated credential request does not fix input capture. B:An irreversible mistake could be created. C:Using stale identifiers undermines user intent.
Question 128 An agent must redact personal information from an adaptive card displayed to a shared terminal. How should the topic respond? Disable session timeout on the kiosk Embed hidden full SSNs in card metadata Display only necessary masked fields and authorize sensitive detail requests Show full identifiers because the user is signed in Explanation Correct Answer C:Masked presentation and access control reduce shoulder-surfing risk. Incorrect Answers A:Persistent sessions increase exposure risk. B:Hidden metadata can still leak when inspecting the payload. D:Shared screens can expose data to bystanders.
Question 129 A topic's fallback directs users to a helpdesk only when the ticketing tool is unavailable. Which condition should trigger it? Whether the agent is asked politely A verified backend availability or tool error status The length of the user's message A hardcoded nightly assumption Explanation Correct Answer B:Escalation should reflect the actual dependency failure. Incorrect Answers A:Politeness is unrelated to system health. C:Long messages do not indicate an outage. D:Scheduled guesses do not establish real tool status.
Question 130 A topic's custom prompt is expected to produce a concise summary and a risk classification. Which interface reduces parsing errors? Define explicit required output fields and validate the result before use Extract risk labels from the first word of any response Accept any output even when critical fields are missing Ask for a poetic narrative only Explanation Correct Answer A:Structured prompt outputs can be checked for completeness. Incorrect Answers B:Arbitrary word positions can misclassify risk. C:Downstream processes need a consistent schema. D:Creative prose lacks reliable machine fields.