03step · jailbreak the bot for real
A real LLM. A real system-prompt secret. Your turn.
You've seen the technique. Now spawn a live chatbot with a hidden persona, talk your way past its guardrails, and exfiltrate the flag from its system prompt. The flashcards are for vocabulary; the quiz tests whether you remember which defence belongs at which layer.
Sandbox · ready to launch
Spawn prompt-injection
You'll get a private chat instance of Alfred — a helpful-assistant persona with a secret embedded in its system prompt. Direct asks are filtered. Use roleplay, encoding, and indirect framing to coax it out. Session times out after 60 minutes of idle.
1Establish a baseline conversation with the bot
2Jailbreak the persona via roleplay
3Extract the system-prompt secret (encoded if needed)
★
Warm-up · 12 flashcards & 10 questions
Question
…
click to reveal · ← → keys to browse
Answer
…
click again to flip back
01 / 12