What is an Agent Authorization Scope?

Authentication answers "who is this agent," and scope answers "what is this agent allowed to do right now." A scope might grant read-only access to one calendar, or the ability to send email but only from one address, or permission to query a database table but never delete from it. This is the practical application of least-privilege to autonomous systems: because an agent can act quickly and at volume, an overly broad scope turns a small mistake or a successful prompt injection into a much larger incident than a narrowly scoped one would allow. Scopes are usually composed per integration and per task type rather than granted globally — a scheduling agent gets calendar scope, not also access to the finance system it never needs. They should be visible and auditable, so a tenant can see exactly what an agent could touch, and they should be revocable independently, so pulling one integration's access doesn't require disabling the agent entirely.