What it does

ECC describes itself as an agent harness performance optimization system. It gives a coding agent a coordinated engineering process of planning, testing, implementing, reviewing, verifying and remembering, so that process does not have to be rebuilt in every prompt. It installs once and ships specialised agents, skills, commands, hooks, rules and persistent memory. It also includes AgentShield, which scans prompts, hooks, MCP configuration, permissions, secrets and agent files. The page lists 71 agents, 311 skills and 97 legacy command shims. ECC works best with Claude Code today and has native plugin paths for Codex and Qoder. Cursor, OpenCode, Gemini, Zed, GitHub Copilot, Antigravity, Qwen and other harnesses get capability-limited adapters, so feature parity varies. Setup runs through a guided npx command or Claude Code's native plugin commands and needs Node.js 18 or newer. The project is MIT-licensed. A hosted GitHub App called ECC Pro covers private repositories.

Capabilities

Specialised agents for planning, review, build repair, security and architecture
Skills covering test-driven development, research, security, docs, frontend, data and operations
Hooks and memory with session summaries and continuous learning
Rules you pick by language or project
AgentShield scanning of prompts, hooks, MCP config, permissions and secrets
Guided setup for Claude Code, Codex and Kimi Code

When to use it

You use Claude Code or Codex and want a ready-made plan, test and review workflow instead of writing your own prompts
You want memory and reusable skills that carry across coding sessions
You want to scan agent configuration for risky permissions or exposed secrets