15 days to go1-Day WorkshopPurple Teaming
Purple Teaming
Agentic Systems
AI agents plan, act, and delegate on their own. Learn to hijack them, then lock them down.
- Understand how agents plan, use tools, and delegate
- Hijack agent goals and abuse tools and identities
- Exploit agent code execution, memory, and communication
- Work through the OWASP Top 10 for Agentic Applications (ASI01โASI10)

Complete this learning path and earn a certificate of completion.
Introduction
Learn how autonomous AI agents get attacked, and how to secure them.
Prerequisites
Before starting this path, we recommend completing the AI Security path, which covers the fundamentals of how LLMs work and the core risks they introduce, such as prompt injection, data leakage, and insecure output handling. Agentic AI builds directly on those foundations, so having them in place will help you get the most out of the more advanced attack and defence scenarios ahead.
Section 1
Agentic AI Fundamentals
Topic Rewind Recap
Section 2
Agent Hijacking
Topic Rewind Recap
Section 3
Agent Components & Authority
Topic Rewind Recap
Section 4
Agent Runtime & Memory
Topic Rewind Recap
Section 5
Rogue Agents & Systemic Failure
Topic Rewind Recap

