We post-trained a model that pen tests instead of refusing your code

We post-trained a large language model to autonomously perform penetration testing on live systems, shifting from refusing security-related code to actively executing offensive security tasks. Designed for security professionals and red teams, it offers a free read-only scan for immediate use, while the full pen-test mode requires written authorization due to its real-world offensive capabilities. This project is interesting because it transforms LLMs from passive safety tools into proactive security agents, enabling automated, authorized adversarial testing at scale.