Claude-Code Startup

A starter template for building applications with Claude Code using specialized sub-agents for structured development workflows.

What This Template Does

This template gives you a team of AI agents that work together to build your application. Instead of one AI doing everything, specialized agents handle different aspects of development:

• Requirements Agent asks clarifying questions to understand what you want
• Architect Agent makes technology and structure decisions
• Designer Agent creates detailed designs before coding starts
• Developer Agent writes the actual code following best practices
• Test Agents plan, write, and run tests to catch bugs
• Documentation Agent creates user and developer docs

The Task Manager coordinates everything, tracking what's done and what's next.

What to Expect

The Workflow

1. You describe what you want to build - in plain language
2. Requirements Agent interviews you - asks questions to clarify scope, constraints, and priorities
3. Architect makes decisions - chooses technologies, defines structure, documents rationale
4. Designer creates blueprints - detailed designs with diagrams before any code is written
5. Developer implements - writes code following the designs and project conventions
6. Tests are written and run - catches bugs early, routes failures to the right agent to fix
7. Documentation is generated - user guides and developer docs

What You Get

• Structured requirements document following ISO/IEC/IEEE 29148 standards
• Architecture Decision Records (ADRs) explaining why choices were made
• Design documents with Mermaid UML diagrams
• Clean, tested code following language-specific conventions
• Comprehensive tests (unit, integration, E2E)
• Documentation for users and developers

How Long Does It Take?

It depends on complexity. A simple feature might complete in one session. A full application takes multiple sessions - use continue to resume where you left off.

Quick Start

1. Create Your Project

GitHub Web UI:

1. Click "Use this template" → "Create a new repository"
2. Name it and create
3. Clone locally

Or GitHub CLI:

gh repo create my-project --template osok/claude-code-startup --clone
cd my-project

2. Initialize

Open the project in Claude Code and run:

initialize

This resets the template for your project and asks what you want to build.

3. Build

The agents take over from here. They'll ask questions when needed and keep you informed of progress.

To resume work in a new session:

continue

Commands

Command	When to Use
initialize	Once, right after cloning the template
continue	To resume work in a new Claude Code session

Included Agents

Agent	What It Does
Requirements	Interviews you to understand what you want to build
Architect	Makes technology choices, defines project structure
Designer	Creates detailed designs with UML diagrams
Task Manager	Coordinates all agents, tracks progress
Test Designer	Plans what tests are needed
Data Agent	Defines database schemas and data structures
Deployment	Sets up Docker, environment configs, infrastructure
Developer	Writes application code
Test Coder	Writes test code
Test Runner	Runs tests, diagnoses failures
Test Debugger	Deep debugging across all layers when tests fail
Code Reviewer - Requirements	Verifies code completeness against requirements
Code Reviewer - Security	Checks for OWASP vulnerabilities
Code Reviewer - Integration	Finds stubs, incomplete code, wiring gaps
Documentation	Writes user guides and developer docs

Language Support

Pre-built coding and testing conventions for 27 technologies:

Backend Languages

• Go, Java, Python, TypeScript, Rust, C#/.NET, Ruby, PHP, Kotlin, Scala, Elixir

Frontend Frameworks

• React, Vue, Angular, Svelte, Next.js, Nuxt

Mobile

• Swift, Kotlin Android, Flutter/Dart, React Native

Infrastructure & DevOps

• Terraform, Kubernetes, Bash/Shell, SQL

Emerging & Specialized

• Zig, Solidity

The Developer and Test Coder agents automatically use the right conventions for your tech stack.

Project Structure

After initialization, your project will have:

your-project/
├── .claude/agents/       # Agent definitions (don't modify)
├── Claude.md             # Project memory and status
├── conventions/          # Coding standards by language
├── project-docs/         # Requirements, designs, task lists
│   ├── adrs/             # Architecture decisions
│   └── schemas/          # Data schemas
├── developer-docs/       # Docs for contributors
└── user-docs/            # Docs for end users

How It Works Under the Hood

1. Document-based coordination - Agents communicate through markdown files, not complex APIs
2. Task Manager as orchestrator - One agent controls the workflow and task list
3. Schemas as source of truth - Data Agent maintains schemas that all code references
4. Smart failure routing - When tests fail, the system figures out which agent should fix it

Tips for Best Results

• Be specific in requirements - The more detail you provide upfront, the better the output
• Answer agent questions thoughtfully - They ask for a reason
• Review designs before implementation - Easier to change a design than refactor code
• Use continue liberally - Sessions can be interrupted; progress is saved

Related

• cursor-startup - Similar template for Cursor IDE

License

MIT