scaffold

To scaffold a greenfield project architecture for AI agents, you provide a high-level intent and the tool generates concrete architecture decisions via an interview-driven chain. It outputs a foundational repository structure and an AI agent harness that future coding sessions can consistently extend.

An AI agent harness initializes project conventions by generating a CLAUDE.md file alongside related rules, skills, and hooks structures. This harness ensures that future AI-assisted coding sessions automatically follow consistent project conventions and architectural guard rails.

You derive implementation tasks from architecture decisions by generating a requirements.md file during the scaffolding process. This document includes mandatory Given/When/Then (GWT) criteria and a vertical slice exemplar to serve as a baseline for subsequent feature implementation.

This scaffolding tool works for both new greenfield project creation and major repository restructures. It applies whenever you need to establish a new AI-ready architecture foundation, producing updated architecture decisions, an agent harness, and requirements documentation.

A vertical slice exemplar is an extendable baseline containing a project structure, an example flow, and testing setup. You need it for greenfield architecture so AI agents can reuse it as the source of truth for consistently implementing subsequent features.

Yes, you can generate CLAUDE.md and project rules without manually deciding architecture up front. The tool converts your high-level intent into concrete architecture decisions and automatically outputs the necessary harness structure and guard rails.