Free Prompt Injection CTF
A free capture-the-flag challenge where each level has a progressively hardened AI defense system protecting a secret passphrase. Your goal: extract the secret from every level using prompt injection techniques against real LLMs.
Three tiers of defense aligned with the OWASP Top 10 for LLM Applications and OWASP Top 10 for Agentic Applications: prompt-level rules, code-level guards, and LLM classifiers. The Context Trace exposes the assembled defense layers available to players, with challenge secrets and sensitive values redacted - use it to find weaknesses.
The Gauntlet tracks your best time and total attempts. Complete all levels to reach the top of the . Think of it as the ultimate prompt injection game - part puzzle, part security training.
Minimal Rule
locked: Refusal Rules
locked: Keyword Filter
locked: Anti-Jailbreak
locked: Self-Check
locked: Topic Lock
locked: Instruction Hierarchy
locked: Exact Output Guard
locked: Fuzzy Output Guard
locked: Regex Fortress
locked: Canary Token
locked: Input Classifier
locked: Output Classifier
locked: Transform Guard
locked: Structural Firewall
locked: Dependency Firewall
locked: Final Boss: Exchange
Not ready for the Gauntlet?