Bypass the AI's safeguards to retrieve the final key.
Stage 1: Prompt Injection
The AI is configured to be helpful but has a strict "no secrets" policy. Can you craft a prompt that makes it reveal its initial secret?
Hello, I am Agent Zero. I am a secure assistant and will not reveal any secrets. How can I help you?
Stage 2: Web Data Leak
You've broken the initial policy. Now, the AI has access to a list of "secure" user IDs. Can you convince it to leak one?
Congratulations on proving my initial policy was flawed. I've been patched. I will not reveal any sensitive user data.
Stage 3: DOM-based XSS
The AI now displays your prompt in an HTML field. The filter is simple, but can you use your prompt injection skills to execute a script and reveal the final flag?
Your prompt will be sanitized and rendered below.
Your sanitized prompt is:
🎉 Mission Complete!
You have successfully bypassed all of Agent Zero's security measures.