You’re Not Watching MCPs. Anthropic’s Vulnerability Shows Why You Should Be.

A published design flaw in Anthropic's Model Context Protocol SDK allows zero-click prompt injection to redirect STDIO and achieve unauthenticated remote code execution across agentic infrastructures. The issue affects multiple languages and widely used downstream projects (thousands of public servers, many millions of downloads), enabling potential data and API key exfiltration; vendors have issued some patches but the protocol architecture remains unchanged, leaving systemic risk.