Matches action effects against agent protect declarations.
Protection is action-centric. A dangerous action must be protected whether it
came from a deterministic DSL action/2 handler or from an optional planner
inspecting an LLM reply.
Summary
Functions
Returns the policy protecting an action, if any.
Returns the matching policy in the active route scope.
Functions
@spec protected_by(module(), Spectre.Effect.t()) :: term()
Returns the policy protecting an action, if any.
:confirm_delete = Spectre.ActionProtection.protected_by(agent, effect)@spec protected_by(module(), Spectre.Effect.t(), Spectre.Definition.scope() | nil) :: term()
Returns the matching policy in the active route scope.
Agent protections are global. Skill protections apply only to effects staged by that mounted Skill, preventing two Skills bound to the same concrete action from stealing each other's policy.