Course Details
Number of Days: 4
Attendance: In-Person Only
Details below last updated: July 27/2026
DETAILS above and below are SUBJECT to change.
Description
This hands-on course teaches students how to supercharge offensive security operations by integrating Large Language Models (LLMs) into a bespoke semi-autonomous framework that can reason, plan, tool-use, and execute complex attack chains leveraging the real-time generation of position independent post-exploitation payloads. Students will select their preferred LLM and iteratively "vibe up" a personalized agentic harness tailored to offensive security workflows. By the end of the course, every student will have a unique, living tool that they can continue developing long after the class ends.
The course will also feature a unique Capture The Flag (CTF) environment featuring thousands of bespoke virtual machines. This scale and customization is intended to deliberately overwhelm traditional manual or scripted approaches, vividly demonstrating why modern offensive operations increasingly require AI assistance in order to achieve offensive security operations milestones at scale.
Key Learning Objectives
By the end of the course, students will be able to:
Design, build, and refine an LLM-driven agentic harness for offensive security tasks.
Implement agentic patterns such as ReAct, Plan-and-Execute, tool calling, memory systems, and multi-agent collaboration.
Integrate the harness with common offensive security tools (Metasploit, Cobalt Strike, Nuclei, etc.).
Leverage LLM reasoning for dynamic target analysis, payload generation, detection evasion, and decision-making under uncertainty.
Operate effectively in high-volume, heterogeneous target environments.
Maintain and extend their personal harness post-course.
High-Level Curriculum Outline
Foundations of Agentic Systems – LLM prompting strategies, tool use, memory, and basic agent loops.
Offensive Security Tool Integration – Connecting LLMs to existing tooling and building custom tools.
Advanced Agent Patterns – Multi-agent orchestration, self-critique, planning, error recovery, and long-running operations.
Scaling to the Swarm – Techniques for operating against large, diverse target sets.
CTF – Live exercise against thousands of unique targets.
Target Audience
Intermediate to advanced offensive security practitioners, red teamers, penetration testers, and offensive security engineers who want to stay ahead of the curve. No prior LLM engineering experience is required, as the course is designed to ramp up students.
Hardware/Software Requirements
A laptop or workstation capable of running multiple docker containers (at least 16GB RAM recommended, 32GB+ preferred).
An LLM subscription/API access of the student’s choice (OpenAI Plus strongly recommended for reliability and speed; alternatives like Claude, Grok, or local models via Ollama/LM Studio are also appropriate).
Basic familiarity with command-line tools, scripting (Python/Bash), and core offensive security concepts.
About the Instructor:
TBA