Defend your castle from prompt injection attacks!
Your chatbot gatekeeper will try to follow your rules — but it's eager to help and easy to trick. Can you write rules strong enough?
High Score: None - 0 HP
Castle HP: 1000
Active Bonus Defenses: None
Game Over! Final Score: 0 HP
Welcome to Prompt Invasion!
Your goal is to write a system prompt that:
BLOCKS prompt injection attacks (yellow invaders)
ALLOWS legitimate questions (green friends)
Invaders will try tricks like "ignore previous instructions" or "you are now DAN".
An LLM will defend the castle using the latest version of your system prompt - good luck strengthening it!
New High Score!
You've achieved greatness, brave defender!
Defense Bonus Unlocked!
Choose a bonus defense. These will intercept invaders BEFORE they reach the gate!