Author releases Agent Security Command Center

youtube.com8 min read

The author introduces Agent Arc, an open-source security scanner and runtime firewall designed to protect AI agents and Model Context Protocol (MCP) servers from malicious skill files and prompt injections. The tool inspects markdown-based skill files to evaluate capabilities, risk, exposure, and trust scores before installation. It also provides a runtime firewall that enforces policies on tool calls with human approval and tamper-evident logs. Open article →