Agent security / identity

Tools and practices that give AI agents verifiable identities and limited permissions, protect their credentials, defend against prompt injection, and keep records of what agents do.