What problem does it solve? Stakeholder feedback (questions, ideas, concerns, bugs) gets lost or forgotten between sessions, and agents either ignore it or nag users with unvalidated guesses. This Skill runs a disciplined signal loop: it pulls open signals, grounds them in the actual code, researches them against the outside world, and either promotes validated ones into the roadmap or dismisses refuted ones with an honest recorded reason. ## Core Features & Use Cases - Churn mode: Picks the oldest open signal, grounds it in the codebase with ministr, researches it with serpapi, records a durable enrichment with a confidence score, then promotes, dismisses, or defers it. - Surface mode: Raises at most one high-confidence, relevant, already-researched signal to the user at the right moment, honoring snooze, dismiss, or act-on-it reactions. - Auditable reasoning: Every enrichment links to a recorded think step and real sources, so a signal's verdict is traceable rather than ephemeral chat. - Use Case: A user types "/signals" mid-project; the Skill researches the oldest open feedback item, finds it is already addressed in the code, and dismisses it with the reason recorded — or validates it and creates a roadmap chunk for /roadmap to build. ## Quick Start Ask the agent to run /signals to churn on the oldest open stakeholder signal and report whether it should be promoted to the roadmap.
