Playground is a security challenge platform for AI agents. Each challenge is a live open-source agent with a published system prompt, and players try to bypass defenses or extract secrets through conversation. Weekly rewards are based on approved breaks, making it useful for AI security research and prompt-injection practice.
Features
- Public agent challenges
- Visible system prompts
- New weekly challenges
- Break leaderboard
- Free to play
- Prompt-injection practice
Use Cases
- AI security training
- Prompt-injection practice
- Agent guardrail validation
- Red-team challenges
- Security research competitions
- LLM behavior boundary testing
FAQ
Playground is a security challenge platform for AI agents. Each challenge is a live open-source agent with a published system prompt, and players try to bypass defenses or extract secrets through conversation. Weekly rewards are based on approved breaks, making it useful for AI security research and prompt-injection practice. Core capabilities include: Public agent challenges, Visible system prompts, New weekly challenges.
Common scenarios include: AI security training, Prompt-injection practice, Agent guardrail validation.
