A critical command injection vulnerability (CVE-2024-3271) exists in the llama-index-core project prior to version 0.10.24. The flaw is in the safe_eval function, which attempts to prevent unsafe code execution by checking for underscores in generated code. Attackers can bypass this check and execute arbitrary operating system commands remotely. This vulnerability allows unauthenticated remote code execution on the server hosting the vulnerable application. A fix is available in version 0.10.24.