Manifold Security releases agentaudit to flag agent config code execution

Manifold Security has released agentaudit, an open source tool designed to inspect AI agent configurations in code repositories before developers open them. The tool flags settings that can execute commands, move data, or widen access by analyzing configuration files for tools like Claude Code, Cursor, and Codex. The author argues that accepting repository configurations has become a reflexive security risk that should be incorporated into organizational threat models. Open article →