Constraint Evaluation
The process of checking an attempted action against all applicable governance constraints to determine whether it should be allowed, blocked, or escalated.
Constraint evaluation is the core operation of moment-of-action enforcement. When an action is attempted, the governance system evaluates it against all active constraints to determine the appropriate response.
The evaluation process considers: - Which constraints are active and applicable to this type of action - Whether the actor has the authority to take this action - Whether any constitutional constraints (invariants) are implicated - What the current governance mode is (Open, Standard, Cautious, Guardian) - Whether any relevant context or precedent affects the evaluation
The result is one of: Allow (action proceeds, trace recorded), Block (action prevented, escalation created), or Escalate (action paused pending human review).
Constraint evaluation is designed to be fast (milliseconds) so it doesn't create friction for routine actions that fall within established boundaries.
How Constellation handles this
Constellation's governance gate performs constraint evaluation on every AI agent action. The evaluation engine checks active constraints, considers governance mode, and returns a result in milliseconds.