Building Orienta — a pre-execution AI governance layer that evaluates agent objectives before execution.