Agentic coding starter kit and workflow boilerplate for Claude Code — Build production-ready TypeScript projects with sub-agents, context engineering, and zero context exhaustion.
⚡ This boilerplate is for developers who want to:
Build TypeScript projects faster with AI-driven workflows
Avoid context exhaustion in long AI coding sessions
Standardize team workflows with specialized AI agents
# 1. Create your project
npx create-ai-project my-project
# 2. Install dependencies (automatic)
cd my-project && npm install
# 3. Launch Claude Code and configure
claude # Launch Claude Code
/project-inject # Set up project prerequisites (read every session by AI)
/implement <your feature> # Start building!
💡 First time? Check the Quick Start Guide for detailed setup instructions
🔄 Updating Existing Projects
Keep your project's agent definitions, commands, skills, and AI rules up to date. Run from your project's root directory:
Your source code (src/), package.json, and other project files are never touched.
Protecting Customized Files
If you've customized a file and don't want it overwritten:
# Add to ignore list
npx create-ai-project update --ignore skills project-context
npx create-ai-project update --ignore agents task-executor
npx create-ai-project update --ignore commands implement
npx create-ai-project update --ignore CLAUDE.md
# Remove from ignore list
npx create-ai-project update --unignore skills project-context
Ignored files are preserved during updates. Note that ignoring files may cause version mismatch with other updated components.
First Run on Existing Projects
If your project was created before the update feature, just run npx create-ai-project update from your project's root directory. It will automatically initialize the manifest by detecting your language from .claudelang.
🚀 Beyond Vibe Coding: Why Sub Agents?
As teams move beyond vibe coding, agentic coding — delegating structured workflows to specialized AI agents — is gaining traction as a practical approach. This boilerplate implements that approach with Claude Code sub-agents:
Traditional AI coding struggles with:
❌ Losing context in long sessions
❌ Declining code quality over time
❌ Frequent session restarts for large tasks
Sub agents solve this through context engineering:
✅ Splitting work into specialized roles (design, implementation, review)
✅ Each agent gets fresh, focused context — no exhaustion
✅ Handling large projects without degradation
This works because Claude Code's sub-agent mechanism runs each agent in its own context window. The parent session delegates tasks, and each sub-agent starts with a clean, focused context. Quality checks (lint, type check, test, build) run locally before each commit — not in CI. The feedback loop stays fast, and code is already verified when it's pushed.
Sub agents working together on a TypeScript project
📝 Slash Commands
Essential commands for Claude Code:
| Command | Purpose | When to Use |
|---------|---------|-------------|
| /implement | End-to-end feature development | New features |
| /task | Single task with skill-based precision | Bug fixes, small changes |
| /design | Create design docs only | Architecture planning |
| /plan | Create work plan from design | After design approval |
| /build | Execute from existing plan | Resume work |
| /review | Check code compliance | Post-implementation |
| /diagnose | Problem diagnosis workflow | Debugging, troubleshooting |
| /reverse-engineer | Generate PRD/Design Docs from code | Legacy system documentation |
| /create-skill | Create a new skill through interactive dialog | Adding project-specific rules |
| /refine-skill | Modify an existing skill with quality review | Improving skill accuracy |
Frontend equivalents (/front-design for UI Spec + Design Doc, /front-build, /front-review, /front-plan) and utility commands (/add-integration-tests, /update-doc) are also available.
This boilerplate ships with general-purpose skills that work out of the box. But skills reach their full potential when you tailor them to your project — your coding conventions, domain constraints, and team decisions.
Think of the included skills as a baseline. The next step is making them yours:
/project-inject — Capture project-specific prerequisites (run once, referenced every session)
/create-skill — Create a new skill through interactive dialog
/refine-skill — Improve an existing skill with optimization review
For principles and best practices on writing effective skills, see the Skills Editing Guide.
Validating Skill Effectiveness
Creating a skill is one thing — knowing whether it actually improves agent behavior is another. Skills interact with each other and with project context, so the only reliable way to know is to run with and without the skill and compare results.
rashomon is a Claude Code plugin for this. Use it when you want evidence, not intuition, that a skill change made things better.
🎨 Skills System
This boilerplate provides the principles used in agentic implementation workflows as skills, making them available for reference in everyday tasks as needed.
Without this boilerplate: ~1 week for setup + infrastructure
With this boilerplate: ~2 days to production-ready application
Success Stories
Sub Agents MCP Server — MCP server enabling Claude Code/Cursor CLI as sub agents
⏱️ Initial development in 2 days — test code makes up ~90% of the codebase, now in production
MCP Image Generator — AI image generation via Gemini API
⏱️ Initial development in 1.5 days — complete creative tool with multi-image blending and character consistency
Both were built using the default /implement workflow — no manual agent orchestration needed.
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
