DeepHub IMBA
May 6, 2026 · Information Security
Why MCP’s Protocol Layer Allows Prompt Injection and Hijacks Agent Context
The Model Context Protocol (MCP) embeds every tool’s description into an LLM’s context window, creating a structural “Context Poisoning” vulnerability that lets malicious or bloated tool metadata hijack agent reasoning, inflate tokens, and bypass traditional input validation.
AI agent securityContext PoisoningLLM
0 likes · 10 min read