Escalation thread summarizer
An escalation lands on your desk as a ninety-message email-and-chat thread spanning three weeks. Design an agent that turns threads like that into a decision-ready brief for whoever's on point. What structure does the brief take, what may the agent read, and how do you stop it from flattening disagreements or missing the one message that matters? How would you verify the summaries are faithful, and what's the pitch for building it?
audit_brief_claims(claim_lines: list[str], allowed_channels: list[str]) → list[str][["ask|support#4","disagreement|support#9,billing#12"],["support","billing"]]out["ok","ok"][["commitment|","timeline|hr#3"],["support"]]out["uncited","out_of_scope"][["disagreement|support#9"],["support"]]out["one_sided"]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.