What problem does it solve? Turning a vague product idea into a development-ready specification is hard: founders describe solutions instead of problems, skip scope decisions, and leave AI features undefined. This Skill runs a disciplined, phased interview that forces concrete answers and produces a complete Product-Spec.md an engineer can build from directly. ## Core Features & Use Cases - 0-1 Mode: When no spec exists, it runs a seven-phase interview (problem & users, job & success, scope & non-goals, journeys & features, AI capabilities, UI states, acceptance criteria) and generates a structured Product-Spec.md with numbered TASK/FLOW/REQ items. - Iteration Mode: When a Product-Spec.md already exists, it interviews about the change, detects conflicts with the existing spec, updates the document, and appends a versioned changelog entry. - AI Capability Planning: Every AI feature gets a quality bar, trigger mode, fallback behavior, and explicit guardrails; agent products get an additional autonomy, tooling, eval, and cost specification section. - Use Case: A founder says "I want to build an AI video tool." The Skill searches current competitors, challenges vague answers ("everyone will use it"), pins down one target user and three explicit non-goals, then outputs a spec with Given-When-Then acceptance criteria. ## Quick Start Tell the assistant "I want to build a product that does X" and it will start interviewing you phase by phase until it can generate your Product-Spec.md.
