0xSteph/pentest-ai-agents

What it solves

It transforms Claude Code into a specialized offensive security research assistant. It solves the problem of needing deep, domain-specific knowledge across dozens of different penetration testing areas (like Active Directory, cloud security, and reverse engineering) by providing a library of specialized sub-agents that can be routed to automatically based on the task.

How it works

The project provides a collection of 50 sub-agents that can be installed as a Claude Code plugin or via a shell script. Once installed, Claude Code automatically routes user requests to the appropriate specialist agent. The system is divided into two tiers: Tier 1 agents provide advisory guidance, while Tier 2 agents are execution-capable and can run underlying CLI tools (such as nmap, sqlmap, or BloodHound) directly on the host system.

Who it’s for

It is designed for penetration testers, red teamers, and security researchers who use Claude Code and want to automate or accelerate their offensive security workflows.

Highlights

• Extensive Coverage: Includes 50 agents covering recon, web, AD, cloud, mobile, wireless, social engineering, and post-exploitation.
• Tool Integration: Integrates with over 80 industry-standard security tools, with a db/doctor.sh script to audit installed tools.
• Automated Routing: Automatically directs tasks to the right specialist agent without manual selection.
• Safety Controls: Implements a "scope-guard" to prevent dangerous operations like DoS attacks or unattended worms.
• Comprehensive Lifecycle: Supports the entire engagement lifecycle from planning and threat modeling to exploitation and professional report generation.