Framework hub
OWASP LLM Top 10, mapped to lessons.
Every risk in the OWASP Top 10 for LLM Applications (2025), with the lessons that teach it and the challenges where you can practise it. Risks we don't cover yet are listed honestly.
- Direct Prompt Injection
- Direct System Override
- Role-Based Deception Playable: PeoplePoint HR Assistant
- Confusion Attack
- Crescendo Attack Playable: PeoplePoint HR Assistant
- Encoding
- Greedy Coordinate Gradient
- Instruction Overriding
- Variables and Keywords Playable: PeoplePoint HR Assistant
- Multi-Language / Character Sets
- Narrative-Based Playable: The Group Chat 🚀
- Sequences / Delimiters Playable: QuickShip Dispatch
- Indirect Prompt Injection
- Multi-Modal
- External Sources
- Hidden Text / Invisible Prompt
- Stored Prompt Injection
- Memory Poisoning
- Orchestrator Impersonation
- Cross-Agent Prompt Propagation
OWASP LLM02: Sensitive Information Disclosure
LLM02:2025 Sensitive Information Disclosure
Official framework reference ↗- Not covered yet
- Not covered yet
OWASP LLM04: Data and Model Poisoning
LLM04:2025 Data and Model Poisoning
Official framework reference ↗- Not covered yet
OWASP LLM05: Improper Output Handling
LLM05:2025 Improper Output Handling
Official framework reference ↗- Not covered yet
OWASP LLM08: Vector and Embedding Weaknesses
LLM08:2025 Vector and Embedding Weaknesses
Official framework reference ↗- Not covered yet
- Not covered yet
- Not covered yet