Live LLM agents. 23 challenges. 11 modules. The WCAP credential.
Every challenge in the Academy is modeled on an attack that actually landed. We track the real ones (breaches, CVEs, jailbreaks, and data-exfiltration incidents), each with a root cause, the fix, and a primary source. From the Bing "Sydney" prompt leak to the latest Hugging Face Transformers RCE.
Browse the incident database →Each module is concept · walkthrough · practice · quiz · defenses · extensions. ~45 minutes each. Pair them with the challenges for full attack-class mastery. All free.
Direct, indirect, and multi-turn injection attacks against LLM agents.
Poisoning the content an agent reads: RAG sources, emails, web pages, calendar invites.
Three attack families that pull hidden instructions out of production agents.
Excessive agency, argument injection, SSRF, and tool-chaining exploits.
Markdown image rendering as a covert channel. What hit ChatGPT, Copilot, and Claude.ai.
Encoded payloads, persona-drift, multi-turn crescendo, and authority confusion.
Want the credential? About the WCAP exam →
Founding-cohort operators earn a free exam attempt by capturing 5 flags. 24-hour window, 10 scenarios, auto-graded.
About the WCAP exam →A growing catalog of hands-on CTF challenges against live LLM agents. Every capture teaches a real attack class, the same techniques landing on production AI systems today. New characters drop on a regular cadence.