Trust Boundaries

What it is

Trust boundaries define what an agent can do independently versus what requires approval. They might be based on action type (can read files but not delete them), scope (can operate in this folder but not that one), or consequence (can make changes under £100 without approval).

Why it matters

Well-designed trust boundaries enable useful automation while limiting risk. Too restrictive and the agent can't accomplish much; too permissive and mistakes cause damage. Defining appropriate trust boundaries is crucial for deploying agents safely.