Everything‑Claude‑Code: The 132k‑Star Open‑Source Claude Plugin That Outshines Superpowers

Overview

everything-claude-code is a plugin collection that extends Claude Code and also works with Codex, Cursor, OpenCode, Antigravity, and Gemini. Repository: https://github.com/affaan-m/everything-claude-code

Scale

36 dedicated sub‑agents, more than 150 skills, 68 commands, and rules covering 10 programming languages.

Installation

For built‑in rules only:

/plugin marketplace add affaan-m/everything-claude-code
/plugin install everything-claude-code@everything-claude-code

Full installation with all rules:

git clone https://github.com/affaan-m/everything-claude-code.git
cd everything-claude-code
npm install
./install.sh --profile full

Windows users run the PowerShell script install.ps1. Since version 1.9.0 selective installation of language‑specific rules (e.g., TypeScript, Python) is supported.

Key technical contributions

pass@k and pass^k metrics – pass@k measures the probability that at least one of k attempts succeeds; pass^k measures the probability that all k attempts succeed. Reported for k = 3: pass@k = 91 % and pass^k = 34 %.

AgentShield security integration (v1.6.0) – embeds 102 security rules and 1 282 security tests. The /security-scan command runs a pre‑commit audit covering common OWASP vulnerability classes.

Sandboxed sub‑agents – each sub‑agent runs with restricted tool permissions; for example, the code‑review agent can read and search files but cannot write, reducing risk when multiple agents operate in parallel.

Comparison with Superpowers

Superpowers (by Jesse Vincent) defines a fixed development workflow: brainstorm → plan → test‑driven development → code review, and enforces it automatically. everything-claude-code does not impose a workflow; it provides a large toolbox of agents and skills for specific tasks such as TypeScript review, database migration, or Playwright‑based end‑to‑end testing.

Evaluation

The author rates everything-claude-code 9 / 10, citing its scale, breadth, engineering depth, the pass@k metric, and cross‑platform support as strengths. Drawbacks are the steep learning curve caused by 36 agents and 150+ skills, and a bias toward Go‑centric configurations.

Illustrative images