ProjectArchitect — Productivity skill for Claude Code

Project Architect 2.0

A governance system for building software with Claude Code. It gives a project a permanent plan, a phase-by-phase build process, a rulebook the assistant re-reads every session, and a build history that any future session can pick up cold. You install it once per project with a single instruction and then work in structured phases from idea to shipped code.

It works for a brand new project, an existing codebase that has no process yet, and an older project that was run with a previous chat-based version of this system.

Who this is for

• If you are a senior developer: this is an opinionated project constitution plus a session protocol that keeps an AI coding assistant on rails. It enforces one-task-at-a-time execution, machine-checkable milestones, per-task commits, and a rules registry that grows as the project teaches you things.
• If you are a solo builder or a "vibe coder": this stops the assistant from wandering. Every project gets a written plan, the assistant checks in at the right moments and runs on its own in between, nothing important is ever left only in a chat window, and you can close your laptop and come back next week and it knows exactly where you were.

Requirements

• Claude Code (the CLI, desktop, or IDE version). This system is built for Claude Code specifically.
• Git. The history and recovery model assume the project is a git repository.
• That is all. There is nothing to compile and no dependencies to install. The package is plain Markdown plus two small shell scripts.

---

Install

Step 1: copy the right folder

This repository contains the README you are reading and one folder called `project-architect-2.0/`. **You copy only that inner folder into your project.** Do not copy the whole repo, and do not copy this README into your project.

ProjectArchitect/                 <- this repo. Do NOT copy this whole thing.
├─ README.md                      <-