Changelog draft generator
Writing the customer-facing changelog is the chore everyone dodges, so it ships late and vague. Design an agent that drafts it from merged pull requests and release notes. The catch: half your merges sit behind feature flags customers can't see yet. How does the agent avoid announcing things that aren't actually live, who signs off before anything publishes, and how do you check its accuracy? Then give me the quick pitch for why this small agent earns its keep.
screen_changelog_entries(pr_records: list[str], flag_states: list[str]) → list[str][["PR-101|dark-mode","PR-102|none","PR-103|"],["dark-mode=100"]]out["publish","publish","hold_unlinked"][["PR-201|beta-search"],["beta-search=25"]]out["hold_partial"][["PR-301|ghost-flag","PR-302|dark-mode"],["dark-mode=100"]]out["hold_off","publish"]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.