What problem does it solve? Vague feature requests and half-formed ideas lead to implementations built on wrong assumptions. This Skill interviews the user in structured rounds, mapping every open decision as a design tree, until a shared mental model exists and the design lands in a written spec. ## Core Features & Use Cases - Round-based design interviews: Delivers the current design with provenance-marked decisions ((you, rN), (my call), (open → Qn), (fog)) plus targeted questions with recommended options, so the user answers by number instead of reconstructing context. - Living spec generation: Writes and rewrites a draft spec (agent/tickets/<feature>/spec.md) per round following SPEC-FORMAT.md, covering problem statement, user stories, properties, decisions, testing seams, out-of-scope items, and fog. - Cross-session continuity: Persists unresolved questions as decision tickets so a fresh session can resume grilling where the last one stopped, then hands the confirmed design to ticket slicing. - Use Case: A user says "I want a notification system, maybe with email digests." The Skill fans out breadth-first across delivery channels, scheduling, and preferences, presents rival designs with trade-offs, and iterates rounds until every decision is settled and the spec is confirmed. ## Quick Start Ask the agent to grill you on your feature idea before any implementation begins, for example: "Grill me on my plan to add team workspaces to the app."