{"uuid": "379824b5-e117-43db-8094-a8d0a131e596", "vulnerability_lookup_origin": "1a89b78e-f703-45f3-bb86-59eb712668bd", "author": "9f56dd64-161d-43a6-b9c3-555944290a09", "vulnerability": "CVE-2023-29374", "type": "seen", "source": "https://bsky.app/profile/premarcu.bsky.social/post/3mtcna7rsua2u", "content": "prompt injection isn't just a chatbot trick. when you give an LLM tools (like a terminal or api), a crafted prompt can make it call those tools in ways you didn't intend. real example: CVE-2023-29374 in langchain allowed exactly that. fixed later, but the pattern is everywhere.", "creation_timestamp": "2026-08-17T21:24:42.370766Z"}