prd

Start by defining the problem, success metrics, and constraints, then structure the document with an executive summary, user stories with acceptance criteria, technical specifications, and risks. This Skill automates that process by interviewing you first, then drafting the full PRD.

AI feature requirements should specify tool and API dependencies, an evaluation strategy with benchmarks, and measurable output quality targets such as precision scores. The Skill includes a dedicated AI System Requirements section covering tools, testing, and validation.

A complete PRD includes an executive summary with success criteria, user personas and stories with acceptance criteria, non-goals, technical specifications covering architecture and integrations, and a risk analysis with a phased rollout roadmap.

Vague terms like "fast" or "easy to use" cannot be tested or verified, leading to misaligned expectations. Requirements should use concrete criteria such as response times under 200ms or a 100% accessibility score so teams can objectively confirm completion.

Discovery should never be skipped, because writing a PRD without clarifying the core problem, success metrics, and constraints produces documents based on assumptions. Asking at least two clarifying questions first prevents hallucinated constraints and scope errors.