What problem does it solve? Turning a finished PRD into a working codebase structure is slow and error-prone: teams over-generate files, mismatch the auth provider, miss schema indexes, or lose traceability between specifications and code. This Skill automates Stage 3 of the Transmute Plan Casting pipeline by coordinating a multi-agent team that scaffolds the entire product — every screen, API endpoint, and data entity — directly from the PRD. ## Core Features & Use Cases - Multi-Agent Code Generation: Spawns five specialized teammates covering backend schema and functions, frontend pages and routing, UI components and design tokens, feature flags and config, and testing plus CI/CD. - Full PRD Traceability: Every generated file carries @traces comments linking back to PRD screen specs, API specs, and data model entities, with a scaffold manifest mapping components to pages and hooks to backend functions. - Prerequisite Gating and Recovery: Validates PRD, BRD, tech-stack.md, spec validation reports, and credentials before running, and supports session recovery to resume interrupted scaffolding without regenerating complete files. - Use Case: After completing Stages 0-2B of the Transmute pipeline, run Stage 3 to produce a development-ready Next.js/Convex (or stack-specific) codebase with tests, CI/CD workflows, ARCHITECTURE.md, and populated CLAUDE.md rules, ready for Stage 5 feature implementation. ## Quick Start Ask the agent to generate the project scaffolding from the completed PRD in plancasting/prd/ using the tech stack defined in plancasting/tech-stack.md.
