Resolves conflicting AI agent memories before they cause bad actions.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.