A critical design-level remote code execution (RCE) vulnerability has been identified in the Model Context Protocol (MCP), a widely adopted open protocol for AI agents, according to OX Security. The flaw, affecting the default behavior of Anthropic's official SDK, allows attackers to execute arbitrary commands on systems using vulnerable MCP implementations, potentially compromising user data and internal databases. The vulnerability impacts multiple programming languages, including Python, TypeScript, Java, and Rust. OX Security highlighted that the vulnerability arises from the STDIO transport method, which enables local process communication. The official SDK's StdioServerParameters can spawn child processes using command parameters from configurations, making unsanitized user inputs executable. Despite the severity, Anthropic has refused to alter the protocol, maintaining that the STDIO execution model is a secure default design. While some vendors have issued patches, the core issue remains unresolved, leaving MCP services at risk of exploitation.