What it does

ECC is a sprawling collection of agent rules, skills, hooks, and MCP configurations meant to make AI coding assistants (Claude Code, Cursor, Codex, OpenCode, etc.) behave more reliably and productively. It ships as a repo you clone and install into your harness of choice, plus a Tkinter dashboard, a Rust control-plane prototype (ecc2/), and a paid GitHub App for private repositories.

The interesting bit

The scale is the spectacle: 249 skills, 63 agents, 79 legacy command shims, 12 language ecosystems, and a one-person maintainer shipping weekly. The project treats “agent harnesses” as a platform layer worth optimizing—token budgets, memory persistence across sessions, verification loops, even sandboxing and CVE scanning via “AgentShield.” Whether that optimization actually moves the needle is left to the user to measure.

Key highlights

• Cross-harness support: Claude Code, Cursor, Codex app/CLI, OpenCode, Gemini, Zed, Copilot
• Session hooks with runtime profiles (ECC_HOOK_PROFILE=minimal|standard|strict) and SQLite-backed state store
• “Continuous learning” pattern extraction from sessions into reusable skills
• Rust-based ECC 2.0 alpha in-tree with dashboard, sessions, status, daemon commands
• MIT-licensed core; $19/seat/month Pro tier for private repos and GitHub App

Caveats

• The 204K stars and 28K forks are self-reported in the README; the repo badge shows the same numbers, so they are at least consistent, if eyebrow-raising for a configuration/rules project
• The “guides” are Twitter/X threads, not in-repo documentation—you are outsourcing your onboarding to a social platform
• The Rust control plane is explicitly alpha and “not yet a general release”

Verdict

Worth a look if you are deep in the daily friction of AI-assisted coding and want a battle-tested (if sprawling) opinionated system. Skip it if you want something minimal, well-documented in-repo, or are skeptical that 249 skills beats a tight dozen you wrote yourself.