All agents have awful security. Mine isn't vibecoded. You might have seen my post about OpenLumara... i challenge you all to hack my public instance of it!
A user has set up a public instance of the OpenLumara AI agent on Discord, running local models and challenging experienced hackers to breach its security. The instance is configured with most modules enabled and includes defenses against common hacking attempts, aiming to test the robustness of OpenLumara's sandboxing and arbitrary code execution prevention. AI
IMPACT Tests the security and potential vulnerabilities of AI agents, encouraging developers to improve defenses against exploitation.