This image is widely shared on social media.
Did they really do that?
The photo seems real because there is what appears to be Hebrew letters in the background and the uniforms look right.
TL;DR: No, the image is photo-shopped.
This is shown by:
Finding an earlier published image which is extremely similar, aside from the content of the placard being Israeli flag (star of David) instead of a message; and no other changes aside from minor image processing ones (different compression ratio, minor cropping)
Preponderance of circumstantial evidence
Image processing evidence (image differences are consistent with photo editing being applied to produce the OP's Twitter image)
Lack of evidence that Twitter image originated from an Israeli soldier despite 200+ reposts.
I did a Google Image Search on this image.
Of the 200+ hits, NONE of them were from an Israeli or pro-Israeli website. The image resides exclusively in Arab or anti-Israeli social media feeds and blogs
Please note that there are TWO data points here:
most of the image's copes seen on Twitter and FB are dated from August 3d and on. I haven't found a single earlier image (this isn't proof by itself, but a data point for proof in Evidence #2).
NONE of the posts linked to the original source of the image on "Israeli" side to prove its originality. (Circumstantial evidence)
Note that the image is nearly identical, save for placards being Israeli flag instead of the offensive message, and being slightly better quality subjectively (I didn't analyze EXIF data, however).
As of right now, I have no formal proof for "nearly identical", but seeking a good methodology to prove that.
The original has Original Transmission Reference field (aka Job Identifier)... while the Twitter one doesn't; which - while not a proof - is consistent with an image taken with a camera vs. one saved in image editing program. I asked for whether it's a valid conclusion on Photography.SE (Circumstantial evidence)
The overall quality of the presumed-original image is higher. For examples, the people in front of white car on the left have more defined skin tone on their hands, and overall the image looks sharper, despite being smaller file size as per EXIF data.
The twitter image is SMALLER, it's cropped.
The presumed-original, for example, shows a human shadow all the way on the right of the image missing from the photoshop copy (and the second word of the text is fully visible as opposed to being cut off at 1 letter); as well as the headlight of the white car on the left of the image.
What the women are holding are binders (you can see the binding spiral on top of them). Having a binder with a slogan used as a poster is not exactly normal or ever done (whereas putting a country flag on top of a binder is quite plausible). (Circumstantial evidence)
That you can not easily change the posting date of a Facebook image (Evidence #2).
That the 2 images are indeed almost identical, aside from slight cropping and the placard content. (Evidence #2).
One one hand, human brain is known to be very good at assessing such similarity (again, citation needed) and clearly a vast majority of voters agree. On the other hand, that's not formal proof even so, and it's a logical fallacy to rely on "majority of voters agree".
As such, I posted a question on Photography.SE for a way to more formally compare 2 images. If it doesn't get closed as offtopic, I will apply whatever solution gets proposed here as further proof of this assumption.
That it's very hard to "uncrop" an image, or add quality. (Evidence #4/5).
(as opposed to cropping it and reducing quality due to lossy compression).
People don't usually use pre-printed spiral binders to print posters as a rule. (Evidence #6).