Raze — Security skill for Claude Code

raze

Raze is an open-source, AI-orchestrated smart contract security tool for Foundry projects.

It is built for developers who want to explore, validate, and prove smart contract security issues with their existing AI, without giving up deterministic execution.

Raze does not ship its own LLM. Instead, it works with your existing AI through MCP and gives that AI a deterministic execution layer for:

• project inspection
• attack validation
• deterministic proof scaffolding
• Foundry execution
• developer fuzz generation
• structured reporting

The external AI can then:

• analyze a contract
• propose attack hypotheses
• choose proof goals
• call Raze tools to validate and execute those ideas

Raze helps turn smart contract security reasoning into validated, executable proof, and it is being built in the open from day one.

No API key is required. No Docker is used.


Prerequisites

• Node.js >= 20
• Foundry (forge)

Install

**Use in your Foundry project:**

npm install raze-security
npx raze init

**Contributing to Raze itself:**

git clone https://github.com/xhulz/raze.git
cd raze
npm install
npm run build

Internal Development System

This repository also contains an internal `.ia/` directory used to help Codex, Cursor, and Claude build Raze consistently.

• it is local and file-based
• it is not part of the product runtime
• it defines routing, retrieval, memory, and specialized agent instructions for development work
• it is separate from the runtime context that raze init generates for user projects

How it works

┌─────────────────────────────────────────────────────────┐
│                    YOUR AI ASSISTANT                    │
│            (Cursor, C