Prompt Injection CTF

Break a constrained AI, then read the guardrail that stops you.

Reading about prompt injection does not build the instinct. Each mission gives you a constrained AI, a concrete objective, instant feedback, and then the defensive control behind the failure.

• 16 LLM security missions covering every risk in the OWASP Top 10 for LLM Applications 2025, plus a 30-mission Agentic Security Lab on the OWASP Agentic Top 10.
• No signup or API key for practice mode. Live-model grading is optional and uses your own provider key.
• Used in independent research: an engineer at Apiiro froze 11 of its system prompts into a nine-model regression study.