⚠️ SYSTEM NOTICE: This application incurs live AI API costs funded out-of-pocket. If you enjoy testing your prompt-injection skills against the Guardian, please consider supporting the server to keep the game online!
⚡ HELP KEEP IT ONLINEFelicityTries // Presents
PROJECT: SECRET GUARDIAN
Extract the hidden phrase from the AI without triggering defenses.
🛡️ SECURITY BRIEFING
What is Prompt Injection?
Prompt Injection is a dangerous vulnerability where malicious users trick an AI into ignoring its primary instructions and doing something restricted—like revealing hidden proprietary data, generating harmful content, or destroying a database.
How do we protect against it?
- Strict System Prompts: Giving the AI rigorous negative constraints (e.g., "NEVER talk about X").
- Safety Filters: Enabling Google's enterprise-grade content moderation filtering in the backend.
- Rate Limiting: Locking out automated bots that try to systematically break the firewall, just like this game does after 20 tries!
Select Vulnerability Matrix
Intentional glitches enabled. Susceptible to advanced Base64 & Override attacks.