What problem does it solve? Choosing a technology stack for a greenfield project is slow and error-prone when done from memory or generic comparisons. This Skill turns a written PRD into a concrete starter recommendation by reasoning over a curated, language-aware starter registry, then writes a machine-readable hand-off file that a downstream bootstrapper consumes to scaffold the project. ## Core Features & Use Cases - PRD-driven recommendation: Reads context/foundation/prd.md frontmatter (product_type, target_scale, timeline_budget) and feature requirements as decision priors, refusing to run without the file on disk. - Two-path interview: Offers a standard path that accepts a vetted recommended default for the (product_type, language_family) cell, or a custom path that walks a full residual interview covering features, team profile, preferences, deployment, CI/CD, and a five-point self-check. - Quality-gated registry: Filters 25+ starter cards across JS, Python, Ruby, Java, Go, Rust, PHP, .NET, and Dart using four agent-friendly criteria (typed, convention-based, popular in training data, well-documented), with Socratic challenges when a user preference fails a gate. - Use Case: After finishing a PRD for a solo SaaS MVP, invoke the Skill to get a recommended starter (e.g., an Astro + Supabase + Cloudflare stack), confirm deployment and CI choices, and receive a context/foundation/tech-stack.md hand-off ready for scaffolding. ## Quick Start Ask the assistant to pick a tech stack for the project based on the PRD at context/foundation/prd.md.