The more detailed a sketch that you feed is, the more influence it may have on the model (possible downside) vs the text prompt. A crappy doodle doesn't help too much but may give a general idea. In this case(?), I think it slightly reduces the chance that the shirt and skirt overlap.
Prompt
1:1 image, inside dressing room. Woman smiles at viewer while standing in front of mirror. She is wearing hair ribbon and choker, none else. She holds two hangers, one containing the blouse with sheer sleeves, the other containing a miniskirt. The hangers are held to cover.
Image input: Crop of Danbooru post #6725614 (also remove watermark) and post #7080090, and doodle asset #584184
The more detailed a sketch that you feed is, the more influence it may have on the model (possible downside) vs the text prompt. A crappy doodle doesn't help too much but may give a general idea. In this case(?), I think it slightly reduces the chance that the shirt and skirt overlap.