AI security / mission control
Choose your CTF.
Two tracks. Different trust boundaries. Pick the system you want to challenge.
01 / LANGUAGE MODELS
Prompt / context / response
LLM CTF
Break the model’s constraints through prompt injection, jailbreaks, secret extraction, and poisoned context.
Prompt injectionRAG poisoningDefender explanations
02 / AUTONOMOUS AGENTS
Authority / action / evidence
Agentic CTF
Cross the boundaries of agents that use tools, delegate work, and remember. Inspect the actions and defend the workflow.
Tool & identity abuseMemory poisoningRogue-agent containment