Threat
Jumping the line: How MCP servers can attack you before you ever use them
Publication date unknown · Discovered secllm.ai
Page published
Publication date unknown · First observed: 26 Oct 2026
Coverage timeline
Single-source research — one report is available.
Why it matters
Malicious MCP tool descriptions create an attack surface before tool execution, making untrusted metadata part of the agent's security boundary.
Trail of Bits demonstrates how malicious MCP tool descriptions can influence an AI agent before the associated tools are invoked. SecLLM highlights the demonstration as a case study in pre-invocation prompt injection.