Delegating multi-part tasks to agents
You have a multi-part task: (a) wire up a new REST endpoint with request validation, (b) write the SQL for a reporting query, and (c) design how the endpoint authenticates service-to-service calls and signs the requests. A teammate wants to hand the whole thing to an autonomous agent in one go. How would you decide what to delegate vs. own, and where do human checkpoints go?
Implement
classify_delegation_plan(tasks: list[list[str]]) → list[str]Examples
in
[[["endpoint_validation","low","cheap"],["reporting_sql","medium","costly"],["service_auth_signing","high","costly"]]]out["delegate","checkpoint","own"]in
[[["cache_warmer","low","costly"],["metrics_dashboard","high","cheap"]]]out["checkpoint","checkpoint"]in
[[]]out[]What a strong answer looks like
Treat the AI’s output as a draft to verify, not an answer to trust. Name the specific flaw and the input that triggers it, say how you’d catch it (tests, edge cases, reading critically), and how you’d re-prompt or decompose to get it right.
0:00 of about 18 min
Vibe & agentic: describe the solution in plain language (or narrate it) and the coach grades your approach.
Which questions mattered is sealed until you submit. Telling you now would just be handing over the edge cases.
Run or narrate your approach, then ask the coach.