Agent Development

NCP-AAI · 31 questions

• A permitting bot must cite ordinance section IDs before advising residents on setbacks. Which prompt-engineering practice most improves that reliability?
• A 311 routing agent must consider today’s street closures before choosing a crew queue. How should dynamic prompt chaining support that?
• Housing staff notice empathy and next-step structure drift across languages in the assistance chatbot. What prompt-engineering fix best improves reliability?
• A clerk agent’s prompt chain fails because step 2 expects fields that step 1 never produced. What development practice prevents that failure mode?
• A library FAQ agent over-answers legal and medical questions outside published services. Which prompt practice keeps it inside service boundaries?
• Residents submit pothole photos that must become structured work orders. How should the city integrate generative and multimodal models in the agent workflow?
• A 311 phone line converts speech to text before an LLM agent triages the call. Where should audio models fit in that multimodal pipeline?
• Building inspectors upload annotated photos, and the agent must not invent defects that are not visible. Which multimodal reliability practice should developers enforce?
• Parks marketing wants one agent stack to draft event captions and generate flyer images. When is integrating generative multimodal capability the right development choice?
• Transit wayfinding agents must accept typed addresses and selfie-at-stop photos. How should developers design multimodal input handling?
• A benefits agent must query eligibility with validated case IDs. What custom-tool design best supports safe municipal system interaction?
• Permit fee amounts must stay accurate for residents. Which tooling pattern should the permitting agent follow?
• Emergency operations wants an agent tool that pages on-call staff through the city’s paging gateway. What is the critical development requirement for that tool?
• Resident-facing agents need GIS layers from the open-data portal but must not alter authoritative maps. How should tool permissions be designed?
• A clerk agent must use three legacy SOAP services for license renewal. What integration approach best fits custom tool development?
• During a storm, GIS lookup times out for a flooding advisory agent. Which error-handling approach is most appropriate?
• The permitting agent’s payment tool receives HTTP 429 responses from the cashiering service. How should error handling respond?
• Eligibility-service failures keep repeating, so residents need a clear fallback instead of endless retries. Which pattern should the benefits agent implement?
• A permitting workflow finds the parcel successfully but the fee service fails. How should the agent handle that partial multi-tool success?
• A GIS address lookup fails with a DNS timeout during a brief network blip, and the agent replies that the street address does not exist. What error-handling approach should developers apply?
• Residents watching a streamed 311 answer can tap “wrong street” while tokens are still arriving. How should the conversation flow be developed?
• Counter staff need to watch tokens stream while an agent drafts a denial letter so they can co-edit live. What development choice best supports that workflow?
• A housing chatbot should ask clarifying follow-ups when retrieval confidence is low. How should its dynamic conversation flow behave?
• Transit chat freezes until a complete answer is ready, and product owners want first-token latency SLOs. Why develop streaming conversation flows?
• An emergency FAQ agent is mid-stream answering a routine question when a higher-priority alert banner must take over. How should streaming be developed?
• After wrong triage of mold complaints, the city reviews which decision heuristic the agent used. What should developers do next?
• Two business-license routing strategies are candidates for production. How should the city refine which decision-making approach to keep?
• Inspectors report the agent over-prioritizes speed versus accuracy on safety tickets. How should decision strategies be refined?
• A grants agent preferred general web search over the city’s official policy API for award rules. How should its tool-selection strategy be refined?
• School-district agents escalate routine questions to humans too early, flooding counselors. What decision-strategy change is appropriate?
• A post-incident review finds the city agent gave legal advice because it lacked a refuse/escalate branch for out-of-scope questions. What decision strategy should be added?