Map agent audit log schemas from actor identity, tool call, data touched, decision reason, and outcome