What problem does it solve? Turning a rough feature idea into a specification that engineers can actually build from is slow and error-prone: requirements stay vague, acceptance criteria go missing, edge cases are forgotten, and non-technical stakeholders cannot read the result. This Skill runs a structured spec-first workflow that researches the domain, challenges the draft from PM/architect/QA perspectives, and delivers implementation-ready documents. ## Core Features & Use Cases - Dual-version specification output: Produces a full technical spec (data models, state machines, API design, error handling, SVG architecture and UI diagrams) plus a plain-language version with zero technical jargon for non-developers. - Research gate before writing: Searches key concept definitions, competitor products, and similar GitHub repositories, then presents comparisons in code blocks for user confirmation before drafting the final spec. - Staged agent development plans: Splits the build into vertical-slice stages (Stage 0 scaffolding, penultimate testing stage, final documentation stage), each with ready-to-paste Codex Instructions and Claude Code Instructions including DoD and rollback notes. - Use Case: You have a one-line idea for a PDF-upload AI summarizer tool. The Skill first shows competitor and open-source comparisons, then delivers the technical spec, a jargon-free spec for stakeholders, and per-stage instructions you can paste directly into Codex or Claude Code. ## Quick Start Ask the AI to turn your rough product idea into a complete technical spec, a plain-language spec, and a staged Codex and Claude Code development plan, noting any constraints or assumptions it should make.
