JFrog Agent Guard

JFrog adds semantic scanning of markdown, scripts and MCP servers to block malicious AI agent behavior

The company's swampUP 2026 release reads skills files and instruction sets for malicious behaviour and lets an organisation cut agents off from public registries at the network layer, which turns agent conduct into a procurement question.