# pentest-ai-agents

A collection of 50 specialized Claude Code subagents that turn Claude into an offensive security research assistant for authorized penetration testing, covering reconnaissance, web, cloud, mobile, exploit development and reporting.

pentest-ai-agents is a set of 50 Claude Code subagents, each carrying deep domain knowledge in one area of offensive security research: network and web reconnaissance, Active Directory, cloud, mobile and wireless targets, social engineering, payload crafting, reverse engineering, exploit chaining, detection engineering and digital forensics. Installing the agent files and describing a task lets Claude route the request to the right specialist automatically.

A recent release added agents for AI attack-surface mapping, code auditing, cryptography analysis, password auditing, and a post-exploitation set covering evasion, persistence, lateral movement and data exfiltration; every offensive agent is paired with the detection technique it exercises. A hardened, least-privilege CI workflow validates each agent's frontmatter and enforces a mandatory scope-guard block on any agent that can run shell commands.

The project installs as a Claude Code plugin or through a shell script, needs no servers or Python dependencies, and is intended for authorized penetration testing, bug bounty and research engagements. It is released under the MIT license.

Website: https://pentestai.xyz/