How are you handling security for AI agents that use MCP tools?
Curious how others are thinking about approaching agentic security for MCP connected agents. External tools processing raw data from websites or other APIs with no sanitization.
There is no distinction between system instructions and whatever the too…