What problem does it solve? Users often bring vague ideas ("I want an app that...") rather than specifications, and typical AI output is a skeleton full of TODOs that never actually runs. This Skill turns a raw concept into a working application by making and naming the undecided decisions, choosing the smallest adequate stack, and verifying the result actually executes. ## Core Features & Use Cases - Concept fixation: Defines who uses the product, the central action, and the hard constraint before writing code, asking everything at once only when truly undecidable. - Justified stack selection: Picks the minimal stack that satisfies the hard constraint (single-file HTML/JS, component framework, or backend with database), with a decision matrix in references/escolha-de-stack.md. - Inside-out construction: Builds model, use cases, boundary, persistence, then interface, so each layer tests the previous one. - Visual identity method: Derives palette, typography, and layout from the subject matter using references/identidade-visual.md, avoiding generic AI-generated aesthetics. - Verified delivery: Requires running the app, exercising error paths, checking narrow widths, and reporting exactly what was and was not tested. - Use Case: A user says "I need a dashboard for tracking my freelance invoices." The Skill decides on a single-file HTML app with localStorage, builds the full vertical, runs it, and delivers the exact command to open it plus a list of decisions made on the user's behalf. ## Quick Start Ask the assistant to turn your idea into a working application, for example: "Create a tool that converts CSV files into charts I can share as a single HTML file."
