@agent

The @agent feature is used for arguments that are agents of their governor at the semantic level in various constructions (e.g., causative, passive, impersonal, etc.).