code-init

Scaffolding a runnable project from scratch requires generating code in an empty or intentionally new directory while enforcing an init-only boundary. This process loads mandatory planning references, produces a structured implementation plan, generates code with file-decomposition constraints, and runs validation with explicit status reporting.

Project initialization cannot modify or refactor an existing codebase. It enforces a strict init-only boundary against existing-repo edits, meaning it applies exclusively to 0-to-1 creation of apps, services, scripts, and sites rather than updates, optimizations, debugging, or extensions.

Architecture planning before generating a new codebase uses an architecture-first workflow that enforces a plan-driven process. It loads mandatory planning references to identify patterns and structure, produces a structured implementation plan, and defines validation steps before any implementation begins.

Frontend initialization supports matching an existing page or site design through an optional frontend reference gate. When asked to match an existing page, it captures reference artifacts and performs reference-aligned visual capture for UI similarity before writing any UI code.

Validating a newly generated codebase after scaffolding involves running validation with explicit status reporting. The scaffolding process completes by generating code with file-decomposition constraints and executing this validation step to ensure the local project is runnable.

Code generation for project initialization should not be used for modifications, optimizations, refactors, debugging, or extensions of existing code. It is designed exclusively for 0-to-1 creation of new apps, services, scripts, and sites in empty or intentionally new directories.