What problem does it solve? Turning a raw feature request into a structured, reviewable product requirements document is slow and inconsistent — use cases get missed, success metrics lack measurement methods, and open questions get silently guessed. This Skill enforces a gated, repeatable PRD authoring workflow. ## Core Features & Use Cases - Use case derivation: Splits a feature into numbered use cases ({FEAT-ID}-UC{n}) with one goal, one actor, one outcome, and P0/P1/P2 priorities. - Measurable success metrics: Requires every metric to include baseline, target, and a measurement method so it can be verified later. - AI feature quality section: For AI features, captures capability, quality bar in user terms, human-in-the-loop policy, tolerated failure modes, and latency/cost ceilings. - Parallel spec writing: Fans out one spec-writer agent per use case when a feature has more than 3 use cases, then merges results. - Use Case: A product owner has an intake document for a new AI summarization feature. Run /prd with the feature ID to produce a gated PRD with UC-IDs, metrics, and an AI quality section, with open questions carried forward as tracked TBD markers. ## Quick Start Invoke /prd with a feature ID, for example: run /prd FEAT-12 to generate the PRD from the existing intake document.
