agent audit logs

AI agent security, identity and permissions

Agent risk now sits in identity and delegated authority. Buyers should approve the credentialed route, not the agent brand, before write access, browsing, code execution or MCP tools move into production.