Idea to Product PRD Generator

Problem

Product ideas often do not arrive as clean specs: scattered descriptions miss users, feature boundaries, and acceptance criteria, while planning docs may mix product behavior with implementation details. Asking a model to draft a PRD directly can produce overcomplete assumptions, unmeasurable requirements, and hardcoded milestones. This skill grades input maturity first, then chooses extraction, targeted clarification, or refusal to generate.

How It Works

• Precheck: If a complete planning doc exists, extract from the PRD summary; if no summary exists, read the full document and tell the user; if only fragments exist, ask at most three questions about product type, core users and MVP features, and tech-stack constraints.
• Serial parsing: Finish input parsing and get user confirmation before drafting the PRD, so later sections do not inherit a wrong premise.
• Parallel references: During drafting, use phase-2-prd-sections.md, prd-template.md, and tech-specs.md to produce functional specs, data models, API or state-management specs, page flows, non-functional requirements, and open questions.
• Quality gate: Before output, verify that user stories have testable acceptance criteria, metrics are measurable, edge states are covered, and open questions are listed; only then deliver the document.

Boundaries

It fits turning rough ideas, MVP plans, or planning documents into a reviewable PRD. It is not a substitute for architecture design, security review, or market validation; when tech stacks or third-party integration details are unclear, the skill should mark them as open questions rather than assuming an implementation path.