# ECC

> Agents, skills, hooks and memory that add an engineering workflow to coding agents

- Repository: https://github.com/affaan-m/ECC
- Category: Agent skills & plugins
- Stars: 286,604 · Forks: 42,513 · Language: JavaScript · License: MIT
- Experience: Some experience
- Page: https://ir.adyog.com/repos/affaan-m/ecc/

A coding agent can write code, but it won't follow a steady process unless you tell it to. ECC packages that process once. The README describes the loop as plan, test, implement, review, verify, remember, improve.

You install it as a plugin (ecc@ecc in Claude Code) using a guided setup run through npx. It adds agents for planning, review, build repair, security and architecture. It adds skills for things like TDD, research and docs. Hooks enforce steps and save session summaries. Rules are optional, and you pick them by language or project.

It works best with Claude Code. Codex has a supported sync path, and the guided installer also covers Kimi Code. Cursor, OpenCode, Gemini, Zed, Copilot and others get adapters with fewer features, so check the README's support matrix first. It also includes AgentShield, which scans prompts, hooks, MCP config, permissions and secrets. The repo is MIT-licensed. A paid hosted GitHub App for private repos also exists.

## What it does

- **Agents for specific jobs**: 68 agents cover planning, review, build repair, security, architecture and domain work. The README says the agent reviews its own work from a fresh context.
- **A large skill library**: 293 skills cover TDD, research, security, docs, frontend, data, ML and operations. 94 legacy command shims remain as entry points while the project moves to a skills-first design.
- **Hooks, memory and learning**: Hooks enforce steps and write session summaries. Continuous learning keeps instincts and turns repeated wins into reusable skills.
- **AgentShield scanning**: Scans agent setups for problems in prompts, hooks, MCP config, permissions, secrets and agent files.
- **Guided multi-harness install**: A wizard sets up Claude Code, Codex and Kimi Code. It shows each destination, checks before writing and asks for one final confirmation.

## What you could build

- Give Claude Code a plan, test and review routine without writing it into every prompt.
- Scan your agent hooks, MCP config and permissions for risky settings with AgentShield.
- Apply language-specific coding rules to an agent across a project.
- Keep session summaries and learned habits so the agent carries context between sessions.
- Set up the same workflow in Claude Code and Codex with one guided install.

## Who it's for

Developers already using Claude Code or a similar coding agent who want a shared workflow. They should be comfortable with Node.js, npx and plugins.

## Get started

You need Node.js 18 or newer, Git and Claude Code 2.1 or newer. Run the guided setup, choose a scope and hook profile (the README suggests Global user and Standard for personal use), then start a new Claude Code session and run /plugin list to confirm ecc@ecc is enabled. Install only from the official sources the README lists.

```
npx ecc-universal@2.2.3 setup
```
```
npx ecc-universal@2.2.3 install --guided
```

Tags: claude-code, agent-skills, hooks, agent-security, codex