AI Agent Security (Course)
Task Scope
Also: scope, action scope
The set of files, systems and effects an agent's task actually justifies, as opposed to everything its permissions let it reach.
In more detail
An agent's permissions describe what it can do; its task scope describes what it should do. Scope errors happen when a valid command resolves to more than intended: an empty variable shortening a path, a wide pattern, a symlink leading outside the project, or a borrowed credential that works on a bigger system. Checking scope means expanding the command to the concrete targets, confirming they lie inside the task's boundary, and asking whether the effect can be undone.