This module aggregates caller-selected AgentToolPolicy.ActionEvaluator findings.
Evaluators run in list order. Enforced blocks outrank flags. Flags outrank allows. A shadow block becomes a flag. The guard retains the first reason at the winning severity.
Summary
Functions
Evaluates an action with the supplied evaluator modules.
Types
@type verdict() :: %{verdict: :allow | :flag | :block, reason: String.t() | nil}
Functions
@spec evaluate(AgentToolPolicy.ActionEvaluator.action(), [module()]) :: verdict()
Evaluates an action with the supplied evaluator modules.