mmlqm/claudesec

ClaudeSec is a shell based security testing framework built for professional red teams, bug bounty hunters, and security researchers doing authorized penetration testing. It uses Claude to read the raw output of well known security scanning tools, decide what matters, and turn scattered findings into a clear report, rather than replacing the tools themselves. The framework follows a defined process. It starts with a reconnaissance stage that runs a chain of established tools in sequence, first finding subdomains, then checking which ones are live, fingerprinting what software they run, discovering hidden directories, and pulling apart their JavaScript for clues. From there, a vulnerability assessment stage runs targeted scanning tools such as SQLMap for database injection, XSStrike for cross site scripting, jwt_tool for token weaknesses, and nuclei for known vulnerability patterns. Findings then go through a manual verification stage, where the AI's suggestions guide a human to confirm real issues rather than acting alone. A key part of the design is reducing false positives and noise. The AI stage groups and prioritizes findings using a modified version of the standard CVSS 3.1 severity scoring system, filters out results it judges unlikely to be real, and looks for ways that several small, low severity issues might combine into one serious attack path. The end result is a report meant for two audiences: an executive summary for stakeholders and a detailed technical version for other security practitioners. The project says it follows recognized industry methodologies including PTES, the OWASP Testing Guide, and NIST's security testing guidance. It runs on Linux, Windows Subsystem for Linux, and Docker, is installed by cloning the repository and running an install script, and is released under the MIT license. The full README is longer than what was shown.