I'm a security researcher who lives at the bottom of the stack. My day-to-day is spent tearing apart binaries, walking Windows kernel internals, and turning AI into a weapon for offensive security.

"If you understand how the machine thinks, you control the machine."

I work across the full offensive spectrum — reading game memory at ring-3, emulating kernel drivers at ring-0, and building AI-native autonomous red-teaming at the agent-orchestration layer. I'm the co-owner of PurpleAILAB and its flagship autonomous red-team agent.

Decepticon is a professional autonomous Red Team agent — and I'm a co-owner of the project. Unlike the "AI hacker" demos that run nmap and print a report, Decepticon executes real attack chains — reconnaissance, exploitation, privilege escalation, lateral movement, and C2 — the way a real adversary would.

It operates under the discipline that separates red teamers from script kiddies: before a single packet leaves the wire it generates a complete engagement package — RoE, ConOps, Deconfliction Plan, and OPPLAN with MITRE ATT&CK mapping — and every action runs inside those rules.

Key capabilities:

• 🧵 Real kill chains — reads an OPPLAN and pursues objectives through whatever path opens up, pivoting and chaining techniques
• 💻 Interactive shells, actually — runs every command inside persistent tmux sessions with automatic prompt detection (msfconsole, sliver-client, evil-winrm)
• 🏝️ Hardened sandbox isolation — all commands run in an isolated Kali sandbox on a dedicated operational network
• 📊 State-of-the-art benchmarks — 102/104 (98.08%) pass rate on XBOW validation-benchmarks

→ Website · Docs · Live App

Project	Description	Stars
Decepticon	Autonomous Red Team agent — real kill chains, RoE/OPPLAN discipline, 98% XBOW pass rate (co-owner)
Vigilo	AI hacker for Web3 smart contracts — bug bounties, audit contests, real-world exploit thinking
omniwire	Infrastructure layer for AI agent swarms — 88 MCP tools, A2A, OmniMesh VPN, CDP browser, 2FA
free-code	Free build of Claude Code — telemetry removed, guardrails stripped, all experimental features	—

Project	Description	Stars
unleash	Unleash Claude Code — 113 patches, zero refuse/telemetry, 42 gates. Bun SEA bytecode patcher + TUI + 11-strategy signature scanner
Kevlar	x64 Windows kernel-driver emulation & behavior-analysis harness powered by Unicorn Engine
vibe-island	A Dynamic Island for AI coding tools — macOS, Windows, Linux (Hyprland/Sway/GNOME/KDE)
AiDA-Fork	AI-powered assistant for IDA 9.0+ to accelerate reverse engineering of C++ games
VoidChecksum	Security research — RE, pentesting, exploit research, security audits
unknowncheats-mcp	MCP server for UnknownCheats & Elitepvpers with automatic Cloudflare bypass
PCILeechGen	Firmware generator tool for PCILeech DMA attacks
pcileech-fpga	FPGA modules for PCILeech Direct Memory Access (DMA) attack software

Offensive Stack: IDA Pro · Ghidra · x64dbg · WinDbg · HyperDbg · PCILeech · Unicorn Engine · Frida · Volatility · Burp Suite · Metasploit · Cobalt Strike

AI / Agent Stack: Claude Code · OpenAI Codex · MCP (Model Context Protocol) · A2A · LangGraph · Agent Swarms · LiteLLM · Prompt Injection & Jailbreak Testing

• 🎮 Deep-diving anti-cheat internals & game engine reversing (Enma / AngelScript automation)
• 🪟 Building kernel-level tooling — driver emulation, manual mapping, hypervisor research
• 🧨 Hunting 0-days across security products & Windows internals
• 🤖 Advancing autonomous red teaming with Decepticon at PurpleAILAB
• 🧊 Expanding PCILeech / DMA / FPGA attack surface

Always looking to collaborate on offensive security research, exploit development, and AI-security projects.

   ┌───────────────────────────────────────────────────────┐
   │  if you understand the machine, you become the machine │
   └───────────────────────────────────────────────────────┘