If you write code with Claude Code but keep retyping the same instructions, manually checking for security issues, and re-explaining context across sessions -- Everything Claude Code (ECC for short) was built for you.

It's a Claude Code configuration system polished by San Francisco developer Affaan Mustafa over more than ten months and crowned champion at an Anthropic hackathon. After being fully open-sourced under MIT in January 2025, it quickly rocketed to 150k GitHub stars, making it the highest-starred Claude Code configuration project.

Project: github.com/affaan-m/ECC

What Is ECC

ECC is not a "prompt collection" — it's complete infrastructure for running AI agents. It decomposes, configures, and automates every step of the development workflow, turning your Claude Code from a chatbot waiting for questions into a self-running digital factory.

The core idea: stop treating AI as a chatbot that waits for your questions, and treat it as the infrastructure of a digital factory.

Core Capabilities

38 Specialized Agents

ECC ships with 38 on-demand specialized agents, each responsible for one specific domain:

• Planner: breaks down tasks and drafts execution plans
• Code reviewer: automatically checks code style and latent issues
• Security reviewer: scans for credential leaks and malicious code injection
• Debugger: locates errors and analyzes logs
• TDD agent: writes tests first, then the implementation

156 Skills + 72 Slash Commands

Skills aren't all loaded at once. ECC has a modular on-demand loading mechanism:

• When you're writing TypeScript, only the TS-specific Review agent activates
• When you start writing Python tests, the TDD agent wakes up
• Frequently used slash commands include /plan, /tdd, /security-scan, /quality-gate, and more

This design keeps 156 skills from simultaneously cramming the Context Window, keeping the system lightweight and agile.

AgentShield Security Defense

ECC ships with 1,282 security tests forming the AgentShield security defense pipeline. Core functions:

• Millisecond-level scanning before the AI executes instructions
• Prevents credentials like API Keys and Tokens from being committed to public repositories
• The --opus flag enables a separation-of-powers mode where the red team (attacker), blue team (defender), and auditor play against each other

Token Optimization

ECC does fine-grained Token optimization at the source level:

1. mgrep replaces traditional grep: filters redundant blank lines and useless information, cutting Token consumption in the retrieval phase by 50%
2. Stop hooks replace UserPromptSubmit: minimizes the latency of saving context memory and completes incremental builds locally

Case Study: A Complete Product in 8 Hours

At the September 2025 Anthropic x Forum Ventures hackathon, Affaan Mustafa and his partner David Rodriguez used this system to build the AI customer research platform Zenith from zero to one in 8 hours.

The workflow has four steps:

1. Input the idea -- describe the product concept and target market
2. AI research & ICP definition -- AI agents research the market automatically and generate the ideal customer profile
3. Synthetic persona interviews -- hold in-depth interviews with AI personas simulating real prospective customers
4. Real user validation -- find real customers and complete the final validation with insight-driven interviews

Claude Code built nearly the entire product; the author says he never typed code by hand. The key to winning wasn't typing speed or on-the-spot prompting, but a complete working system polished in advance.

How to Get Started

# Clone the repository
git clone https://github.com/affaan-m/ECC.git

# Configure your Claude Code following the README
# Put CLAUDE.md in the project root
# Configure .claude/settings.json

ECC covers 12 language ecosystems — whether you use TypeScript, Python, Rust, or Go, you'll find matching skill configurations.

Who It's For

• Developers who use Claude Code every day
• Teams that want to upgrade AI coding from "chatting" to an "engineering process"
• Individual developers who want a security line of defense against AI missteps

References:

• GitHub repository
• Chinese README