Wraith Academy is a hands-on training platform focused on offensive AI security. It teaches how to break and exploit LLM-powered systems through interactive challenges and real-world attack scenarios.
Built around a CTF-style experience, learners directly interact with vulnerable AI agents to extract secrets, bypass guardrails, and understand how modern AI systems fail in practice.
Key Features
- Hands-on CTF challenges Practice against live AI agents with real attack paths like prompt injection and system prompt extraction
- Real-world attack techniques Learn techniques actively used against production LLM systems
- Structured learning modules Covers prompt injection, tool abuse, data exfiltration, jailbreaks, and more
- Progressive difficulty Beginner to advanced scenarios including multi-turn manipulation and RAG poisoning
- Defense-focused learning Each challenge includes explanations and mitigation strategies
What You’ll Learn
- Prompt injection attacks Direct, indirect, and multi-turn exploitation techniques
- System prompt extraction Methods to reveal hidden instructions and configurations
- Tool abuse Exploiting insecure integrations like file access and external APIs
- Data exfiltration Techniques for leaking sensitive information from AI systems
- Guardrail bypasses Jailbreak strategies and alignment evasion
Why It Stands Out
- AI-native offensive training Focused entirely on modern LLM attack surfaces
- Interactive learning model Learn by breaking instead of passive content
- Free core content Full academy access without paywall, optional certification available
- Gamified experience Story-driven challenges with characters and scenarios
Wraith Academy is ideal for security researchers, red teamers, and developers looking to understand and secure AI-powered applications through offensive techniques.
