Tools
LLMVault Ships an Intentionally Vulnerable OWASP LLM Top 10 Range for Agent and RAG Penetration Testing
CyberSunil/LLMVault (263 stars, created 2026-07-15, Python, Docker) is a deliberately vulnerable training platform covering the OWASP LLM Top 10, with tracks for prompt injection, RAG security, agent security, and GenAI penetration testing, tagged for CTF use. It is the defensive-training counterpart to the detection tooling shipping the same month, and one of only two agent-security repos created in the last 30 days that cleared 200 stars. Treat the star count as early signal rather than validation, since the scenario quality of a vulnerable-by-design range is hard to assess from the outside.
Source
↳ Follow the thread