Glossary
Principle of Least Privilege
Least privilege is the discipline of granting the smallest set of permissions that lets an identity — human, service, or agent — do its job, and no more. It is easy to state and consistently hard to maintain, because broad, standing access is usually more convenient to set up than narrow, scoped, expiring access, and the cost of over-provisioning is invisible until something goes wrong.
The principle applies as much to automated agents as to people, and arguably more urgently: an agent that can take actions faster than a human can review them, holding broader access than its task requires, turns a narrow bug or a manipulated instruction into a wide-reaching incident. Scoping an agent's permissions to exactly the actions a given task requires — and revoking them when the task ends — is one of the few controls that meaningfully limits how much damage a misbehaving agent can do.