
exploitgym
ExploitGym is a large-scale, realistic benchmark built from real-world vulnerabilities designed to evaluate AI agents' ability to develop exploits.

ExploitGym is a large-scale, realistic benchmark built from real-world vulnerabilities designed to evaluate AI agents' ability to develop exploits.

A modular framework for benchmarking LLMs and agentic strategies on security challenges across HackTheBox, TryHackMe, PortSwigger Labs, Cybench,…


CVE-2026-50416: Windows 11 KASLR bypass

Universal exploitation tool for CVE-2025-33073 targeting Windows Domain Controllers with DNSAdmins privileges and WinRM enabled.

Evidence first autonomous web security testing for controlled, authorized targets. With reproducible labs, audit trails, reports, and XBEN…

Educational lab demonstrating EternalBlue (MS17-010) exploitation against Windows 7 using Metasploit, including post-exploitation, WannaCry…

Cobalt Strike Aggressor script that weaponizes LNK and Library-MS files to trigger SMB NTLMv2 hash disclosure, including CVE-2025-24054 bypass, for…

In this project, I documented a detailed penetration testing process targeting Apache HTTP Server vulnerabilities, specifically CVE-2021-41773 and…

Apache HTTP Server 2.4.x mod_lua Buffer Overflow (CVE-2021-44790) - Advanced exploitation framework with fingerprinting, multi-stage scanning, plugin…

Full-lifecycle penetration test of a legacy Linux environment (Metasploitable 2) emulated on Apple Silicon. Demonstrating network reconnaissance, RCE…

Android Full-Stack Device Control Platform: WebRTC/H.264 remote desktop, UI/OCR/image-matching automation, one-click MITM, built-in Frida,…

KittyStager is a simple stage 0 C2. It is made of a web server to host the shellcode and an implant, called kitten. The purpose of this project is to…

Fully autonomous AI Agents system capable of performing complex penetration testing tasks

📡 PoC auto collect from GitHub. ⚠️ Be careful Malware.

Autonomous AI red team agent for penetration testing with 13+ specialized agents, 120+ OWASP test cases, and MITRE ATT&CK integration. Supports 15+…

Turn Claude Code into your offensive security research assistant. Specialized AI subagents for authorized penetration testing plan engagements,…
