TL;DR: The history of the texts makes it impossible to be sure when stage directions, or even attributions to speakers, were added to Classical Greek plays. A possible (but doubtful) early candidate for stage directions that were original to the playwright are the noises of the Erinyes in Aeschylus’ The Eumenides (458 BCE).
Attributions to speakers
There’s some doubt, in the case of Classical Greek drama, as to whether even the attributions of speeches to speakers go back to the original texts. We don’t know exactly how Greek actors learned their lines, but it seems possible that they did so by direct instruction from the playwright:
The playwright himself almost always served as stage director, or didaskalos (‘teacher’), and the Greek phrase for directing a play was didaskein choron, ‘to teach a chorus’.
Rush Rehm (1994). Greek Tragic Theatre, p. 25. London: Routledge.
In a system where the playwright is teaching the lines to the actors and chorus, there would be little need for the playwright to annotate each speech with the name of its speaker, as the speaker would be obvious from the dramatic situation, especially in Classical Greek theatre where the number of actors on stage at a time is limited to two or three. And indeed there is evidence that the earliest manuscripts did not name the speakers, but only indicated the end of one speech and the beginning of the next, so that the speaker attributions that appear in later manuscripts are the deductions of copyists.
The 13th and 14th century manuscripts in general preface each new speech in the dialogue with an abbreviated form of the speaker’s name (or description, if he is nameless). This system we have taken over not only for modern editions of classical texts but for all dramatic texts. It has become so universal that it now seems the obvious and natural method and we have difficulty in understanding how any other could ever have been used.
Nevertheless occasionally in [manuscripts A and M] changes of speaker are marked simply by a dash, which is clearly a survival of the paragraphus† common in papyri and attested by Hephaestion. A also preserves traces of the use of the double point (:) to mark changes of speaker. In Γ in one play, the Ecclesiazusae, most of the changes of speaker are marked, if at all, with a double point, and only 24 with the speaker’s name; this was no doubt the least read play in the MS, not furnished with glosses or other aids to the inexperienced reader, and so most likely to preserve archaic features.
J. C. B. Lowe (1962). ‘The Manuscript Evidence for Changes of Speaker in Aristophanes’. Bulletin of the Institute of Classical Studies 9:1, p. 27.
† “A short horizontal stroke drawn below the beginning of a line” (OED).
Another line of evidence for the lack of speaker attributions in the earliest texts comes from disagreements between scholars.
Notes which explain who is speaking, even where in our MSS an attribution is prefixed to the line in question […] confirm that in earlier texts attributions were often lacking. Others give alternative attributions or record the different attributions of previous commentators or texts […]. Differences of opinion are also recorded as to where a change of speaker should be placed, sometimes with a reference to the double point […] or the paragraphus.
Lowe, p. 37.
Stage directions in tragedies
There are a handful of annotations in manuscripts of Classical Greek tragedies that can be interpreted as stage directions. But as with attributions to speakers, there is some doubt as to whether these are original, or were added by later copyists. Taplin (1977) considered all such instances, and found that
of the thirteen candidates for authentic parepigraphai [that is, stage directions] seven are almost certainly the fabrications of pedants, ancient or modern, and so are two others more likely than not. That leaves just four—the Erinyes’ noises in Eumenides, Silenus’ labials in Diktyoulkoi, the singing from Polyphemus’ cave in Cyclops, and the strange ‘songs’ of the strange Cassandra fragment (which does not really count)—just four which might well go back to the dramatist himself. In each case I hope to have shown that this is not the only possible explanation.
All four are, of course, aural; they all direct sounds of one sort or another, whether on or off stage. We might be tempted to argue that, while the dramatists did not add any other kind of stage direction, such as gestures or costumes, they did indicate noises. The great objection to this is that there are many places where noises which are clearly inferable from the text are not indicated explicitly, neither by phonetic transcription nor by a parepigraphe. A few examples are the trumpet at Aeschylus’ Eumenides 573, the lyre in Sophocles’ Ichneutai, the approach of Philoctetes at Philoctetes 201 ff., the distant sound of the murder at Euripides’ Electra 747 ff. So it is surely a much more likely hypothesis that the convention arose that keen readers might, if they liked, write in their own parepigraphai indicating any noises they thought they could infer from the text; especially since it has now been shown that a convention of this sort grew up in the matter of the attribution of parts. This would explain why such parepigraphai survive at random and why some are demonstrably wrong.
O. Taplin (1977). ‘Did Greek Dramatists Write Stage Instructions?’ The Cambridge Classical Journal 23, p. 127.
The fact that there are so few cases in the entire corpus of Classical tragedy, and that all of them are more or less doubtful, is evidence that the tragic playwrights did not write stage directions: after all,
the διδάσκαλοσ was present at the rehearsals of his plays and could teach his stage directions on the spot; so there was no call for him to write them down.
Taplin, p. 129.
Taplin says that the strongest candidates for stage directions that might have originated with the playwright are the noises of the Erinyes (Furies) near the start of Aeschylus’ The Eumenides, where the chorus (representing the Erinyes) are woken by the ghost of Clytemnestra:
Κλυταιμήστρας Εἴδωλον μύζοιτ᾽ ἄν, ἁνὴρ δ᾽ οἴχεται φεύγων πρόσω:
φίλοι γάρ εἰσιν οὐκ ἐμοῖς προσεικότες.
Κλυταιμήστρας Εἴδωλον ἄγαν ὑπνώσσεις κοὐ κατοικτίζεις πάθος:
φονεὺς δ᾽ Ὀρέστης τῆσδε μητρὸς οἴχεται.
Κλυταιμήστρας Εἴδωλον ᾤζεις, ὑπνώσσεις: οὐκ ἀναστήσῃ τάχος;
τί σοι πέπρωται πρᾶγμα πλὴν τεύχειν κακά;
Κλυταιμήστρας Εἴδωλον ὕπνος πόνος τε κύριοι συνωμόται
δεινῆς δρακαίνης ἐξεκήραναν μένος.
Χορός (μυγμὸς διπλοῦς ὀξύς)
λαβὲ λαβὲ λαβὲ λαβέ, φράζου.
Ghost of Clytemnestra Whine, if you will! But the man is gone, fled far away. For he has friends that are not like mine!
Ghost of Clytemnestra You are too drowsy and do not pity my suffering. Orestes, the murderer of me, his mother, is gone!
Ghost of Clytemnestra You moan, you drowse—will you not get up at once? Is it your destiny to do anything other than cause harm?
Ghost of Clytemnestra Sleep and toil, effective conspirators, have destroyed the force of the dreadful dragoness.
Chorus (with whining redoubled and intensified) Catch him! Catch him! Catch him! Catch him! Look sharp!
Aeschylus (458 BCE). The Eumenides, lines 117–130. Translated by Herbert Weir Smyth (1926). London: Heinemann.
But you can see from this passage that even if you only had Clytemnestra’s lines, the moaning and whining of the Erinyes would be clearly indicated by her words:
before jumping to conclusions it should be noticed that any careful reader can work out that there must be noises from the Erinyes in between each couplet, and he could supply the nouns from μύζοιτ᾽ ἄν in 118 and ᾤζεις in 124.
Taplin, p. 122.