Kali Linux Docker + Claude Skill

Token-Efficient Pentesting System for Claude Code

🎯 Overview

This project provides a comprehensive Kali Linux Docker container with 200+ pentesting tools, optimized for use with Claude Code through a custom skill. By using direct command execution instead of MCP protocol, this system achieves 70% token efficiency compared to traditional MCP server approaches.

🏗️ Architecture

┌─────────────────────────────────────────────────────────┐
│                    Claude Code                          │
│                                                         │
│  ┌──────────────────────────────────────────────────┐   │
│  │           kali-docker-pentesting skill           │   │
│  │  (Tool catalog + usage patterns)                 │   │
│  └──────────────────────────────────────────────────┘   │
│                         │                               │
│                         ▼                               │
│  ┌──────────────────────────────────────────────────┐   │
│  │              bash_tool                           │   │
│  │  (Direct docker exec commands)                   │   │
│  └──────────────────────────────────────────────────┘   │
└─────────────────────────────────────────────────────────┘
                         │
                         ▼
┌─────────────────────────────────────────────────────────┐
│            Kali Linux Docker Container                  │
│                                                         │
│  • 200+ pentesting tools                                │
│  • No MCP server overhead                               │
│  • Direct command execution                             │
│  • Persistent storage via volumes                       │
└─────────────────────────────────────────────────────────┘

📦 What’s Included

Docker Container

• Base: Kali Linux Rolling (kept current via in-place apt full-upgrade — see Updating)
• Tools: 200+ pentesting tools
  • Network scanners (nmap, masscan, zmap)
  • Web app testing (nikto, sqlmap, burp suite)
  • Password cracking (john, hashcat, hydra)
  • Wireless tools (aircrack-ng, wifite, reaver)
  • Forensics (volatility, binwalk, foremost)
  • Exploitation (metasploit, SET, beef)
  • LLM / prompt-injection testing (garak, promptmap2, promptfoo, PyRIT, agentic-radar) — see below
  • And much more!

LLM / Prompt-Injection Testing

For assessing LLM-backed applications (chatbots, RAG, AI agents) against prompt injection, jailbreaks, and data leakage — mapped to the OWASP Top 10 for LLM Applications and MITRE ATLAS: