As Schwyzer pointed out in his article on expressions of the agent, the Attic tragedians provide many examples of non-standard agent markers (1943: 20–8). In fact, ὑπό is so rare in comparison to other prepositions, that, were these plays our only source of Ancient Greek, we would not at first glance be able to pinpoint it as the default agent marker. In iambic passages of the Oresteia, for instance, ὑπό only occurs in three PACs, as against seven with πρός+G, two with ἐκ, and one with παρά. Does this variety of agent marking mirror that in prose of the period, or is it simply a feature of poetic diction, conditioned by the meter? In order to answer this question, I will look at prepositional PACs in Aeschylus, Sophocles, and Euripides. To impose some uniformity on the data, I will only consider PACs in iambic passages. Because this limitation reduces the number of PACs under consideration, I will extend the study to those constructions where agent expressions occur with intransitive verbs, such as πάσχω or ἀποθνῄσκω, that act as suppletive passives to transitive counterparts like ποιέω and ἀποκτείνω. Such constructions are quite common in tragedy because of the frequent description of suffering in the genre.
In general, the pattern in prose whereby certain agent markers are associated with certain verbs does not hold true in poetry.