Skip to content
← Tools / Skills
Skill · .md packageaffaan-m/10 trustVerified
ECC logo

ECC

The agent harness optimization system — 278 skills, 67 agents, memory, security, for every major agent.

Visit official site →

About

ECC (by affaan-m) is a complete agent-harness performance optimization system, not just a skill pack: skills, instincts, memory optimization, continuous learning, security scanning, and research-first development, evolved over 10+ months of daily production use. It ships 278 skills and 67 agents covering Claude Code, Codex, Cursor, OpenCode, Gemini, Zed, and GitHub Copilot, with hooks for persistent memory (session-end Stop hooks save context to markdown files; Pre/PostToolUse hooks capture development patterns and promote recurring ones into skills via /evolve). It grew out of an Anthropic hackathon win and is one of the most-starred agent repos on GitHub (~230k stars), MIT-licensed, with a commercial ECC Pro tier ($19/seat/mo) for private repos and a GitHub App.

The honest picture: ECC is enormous — installing the whole system is overkill if you want two or three skills, and the learning curve is real. The harness-heavy approach (hooks, memory files, evolving instincts) adds moving parts that can surprise you when they misfire. The commercial upsell (ECC Pro, GitHub App) sits alongside the MIT core, so check what is free versus paid before committing a team.

What it can do

  • 278 skills and 67 agents across harnesses
  • Persistent memory via hooks and markdown files
  • Continuous learning: instincts promoted to skills
  • Security scanning for agent workflows
  • Research-first development patterns

Setup

Clone github.com/affaan-m/ECC and follow the harness setup docs; works across Claude Code, Codex, Cursor, OpenCode, Gemini, Zed, Copilot.

Prerequisites

A supported AI coding agent; comfort with hooks and markdown-based memory files.

Use cases

coding-agentsagent-harnessagent-memoryagent-security

Works with

Trust score — 9/10

/10 trustVerified

Scored 9/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 50; Trending score 90/100 from researcher assessment.

★ 228,000 GitHub stars

Sources

Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.