This research introduces a novel system for semi-autonomous penetration testing using large language models, addressing the challenges of full autonomy in specialized cybersecurity tasks.

• Combines LLM capabilities with human expertise to execute complex cybersecurity workflows
• Demonstrates practical applications through testing on Hack The Box virtual machines
• Addresses LLM limitations in security reasoning while maximizing their analytical strengths
• Establishes a framework for human-AI collaboration in high-stakes security environments

This research matters because it offers a pragmatic approach to integrating advanced AI into cybersecurity operations, enhancing efficiency without sacrificing the critical human oversight needed in security contexts.

Construction and Evaluation of LLM-based agents for Semi-Autonomous penetration testing