TushaeBXN/kerrigan-fantasma
Custom security LLM on a Recurrent-Depth Transformer. Writes exploit harnesses, fuzzes them, learns from crashes, and reasons across hardware-software boundaries. Trained on Linux kernel, UEFI, CPU specs, CVEs, and 12 languages.
🎯 Vulnerability: Web Security Vulnerability
💡 Defense Fix: Implement strict input validation, contextual output encoding, and enforce least-privilege authorization controls.