shipwright

# Shipwright ⚓

### One skill to design it, build it, test it, polish it, and ship it.

Shipwright is an AI coding agent skill that covers the **complete software lifecycle** — from a rough idea to a production-ready system built to the engineering standards of Airbnb, Stripe, and Meta. 

Drop it into [Claude Code](https://github.com/anthropics/claude-code) or [Cursor](https://cursor.com) and your agent gains the ability to design architecture, implement with TDD, create distinctive frontends, run exhaustive QA across every layer, harden security, optimize performance, and ship with confidence.

---

## Why Shipwright?

Most AI coding agents are great at writing code but terrible at everything else — they skip design, ignore security, produce generic UIs, forget about accessibility, and have no concept of production readiness. You end up babysitting every decision.

Shipwright fixes this by giving your agent a structured methodology that covers everything from "I have an idea" to "it's live and bulletproof." It's the accumulated knowledge of what world-class software actually requires, distilled into a format that AI agents can follow.

---

## What It Does

Shipwright has two modes:

### 🔨 BUILD MODE — For new projects

A design-first workflow that prevents the agent from jumping to code before thinking:

**Think** (failure modes, constraints, requirements) → **Architect** (domain model, service boundaries, data patterns, API strategy) → **Plan** (task breakdown with TDD, YAGNI, small commits) → **Implement** (build with craft, no AI slop) → automatically flows into Ship Mode

### 🚢 SHIP MODE — For taking software live

A 12-phase production-readiness audit that covers every layer:

| Phase | What the Agent Audits & Fixes |
|:---|:---|
| **1. Code Health** | Dead code, dependencies, secrets, type safety, build warnings |
| **2. Backend & Scalability** | API design, database optimization, caching, connection pooling, background jobs, data integrity, circuit breake