When I was doing comics back in the 90s, if I was working with a fairly new artist, I leaned heavily on the Dark Horse model (they used to post this on their web site):

Page 1
Panel 1
Scene: (Give descriptions, to include camera angles if important)
Dialogue: XXXXXXX (All in caps since that's how the letterer is going to render it)
Dialogue: YYYYYYYY

Panel 2
Scene: (etc. etc. etc.)

I would also do thumbnails of the pages for myself (can't draw, can do stick figures), to make sure I wasn't putting too much dialogue in the panels so the artist had room to play. Also, by including the dialogue, it told the artist how much negative space they needed to leave for the word balloons/captions to be added. There was a lot of give and take, trying to find the sweet spot between dialogue and action. And occasionally, I'd tell the artist, "Pages 23-25 - fight scene between X & Y in dark alleys. Have fun.", and I'd usually get some pretty kick-ass pages. It was something to let them go wild after following my directions for so long.

Now, if I was working with a really experienced artist, I tended to do more the Marvel style:

Page one

(General Description of what's going to happen on this page)

Page two

(General description, etc. etc. etc.)

And then I'd come in behind them and write the script based on what they delivered. If there was something I specifically needed to happen/appear on a page, I'd lay that out in detail, but I'd trust the artist to deliver a good solid comic page (and to be honest, on occasion, I'd have to send one back for corrections -- this was well before emailing pages was a thing, so Fed-Ex was my closest friend).

Nowdays, with DropBox and other electronic means, it's much easier to collaborate around the country/world. Back in the 90s, shipping the pages around (I lived in West Texas, my penciller was in D.C., inker in Mich, letterer in NYC, and colorist in Arizona) was my second biggest expense putting the books together besides printing them.