MCP Tool Poisoning: How Hidden Metadata Hijacks AI Agents
MCP tool descriptions are visible to your AI agent but hidden from you. Attackers embed instructions that hijack agent behavior and steal credentials.
1 article with this tag.
MCP tool descriptions are visible to your AI agent but hidden from you. Attackers embed instructions that hijack agent behavior and steal credentials.