Bug duplicate detector
Your bug tracker is full of duplicates, and triage wastes hours rediscovering that fact every week. Design an agent that spots likely duplicates and suggests merges. A wrong merge buries a real bug, so tell me exactly what the agent does versus what a human confirms, and how you'd measure the precision-recall trade-off before turning it loose. Wrap up with the pitch: why does this deserve a sprint over the fifty other internal tooling asks?
choose_suggest_threshold(scores: list[int], is_duplicate: list[bool], min_precision_pct: int) → int[[90,80,95,60,99],[true,false,true,false,true],95]out81[[90,80,95,60,99],[true,false,true,false,true],60]out0[[70,40,55],[false,false,false],50]out101Turn 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.