Survey response analyzer
Your quarterly developer-experience survey collects hundreds of free-text answers, and leadership wants themes. Design an agent that analyzes it. Those answers were written under a promise of anonymity, and some name names or describe situations only one person could have lived — so what guardrails does the agent enforce before any output reaches a manager? How do you verify the themes are faithful and nothing traceable leaks, and how do you pitch this so engineers still trust the survey next quarter?
screen_survey_themes(theme_names: list[str], respondent_counts: list[int], has_identifiable_detail: list[bool], k_floor: int) → list[str][["flaky ci","laptop specs","review turnaround"],[12,7,3],[false,false,false],5]out["flaky ci:release","laptop specs:release","review turnaround:suppress"][["oncall load","manager named in quote"],[9,6],[false,true],5]out["oncall load:release","manager named in quote:paraphrase"][[],[],[],5]out[]Turn the fuzzy goal into a bounded agent task. Say what the agent reads, what it may do on its own versus draft for a human, and what it must never touch. Then earn the trust: how you’d verify it before rollout, and the pitch (the problem, what you built, why it matters).
Agent build: design the agent in plain language (or narrate it). The coach grades the decomposition, the guardrails, the verification plan, and the pitch.