"Working spaces" in PS means the default profile - it can be overridden by profiles embedded into images. As you have "preserve embedded profile" set, and "ask when opening" not selected, when PS finds an embedded profile it switches to it automatically, without asking.
RAW images could have their own specific color space, or could use one of the standard ones (it's up to the camera maker). If it's a specific space, ACR (and LR) turn it into one of the standard ones (using its standard camera profiles, or you can create your own for extreme accuracy).
Anyway, in ACR you can set what space the RAW image will be converted to when opening it in Photoshop:
Learn how to work with color spaces in Camera Raw.
helpx.adobe.com
In this chapter from <a href="http://www.peachpit.com/store/photoshop-cc-visual-quickstart-guide-2014-release-9780133980462?w_ptgrevartcl=Camera+Raw_2247674">Photoshop CC: Visual QuickStart Guide (2014 release)</a>, you’ll learn about the Camera Raw tools and tabs; choose workflow options for...
www.peachpit.com
If you want to work in ProPhoto RGB, you should select it in ACR, and not just convert, say, an AdobeRGB image into ProPhoto RGB in PS later. The reason is that as ProPhoto RGB is "larger" than AdobeRGB, some colors could have been clipped when the RAW -> AdobeRGB conversion happened (and cannot be restored later), while RAW -> ProPhoto RGB could have kept them. That depends on the camera capabilities, though, it has to be able to record colors outside AdobeRGB. Many modern cameras can also generate images that won't fit well in 8 bit per color channel.
Still, there are good reasons to work in a larger color space (and higher bit depths, i.e. 16 bit), even when the image is in a smaller one.
When you edit an image, the various pixel values will be shifted. Whenever they fall outside the boundaries, they need to be remapped to the nearest allowable value More edits, more of that, and more unrecoverable losses. The larger the space, less risks of that happening. Of course, a lot depends on how many values in the images are already closer to the boundaries, and how "strong" the edits are. With some images and edits the losses could be quite minimal.
Then, only the "final" image needs to be converted into a smaller space/bit depth - i.e. 8 bit sRGB for web display, or a printer-specific space - so any loss is reduced to the necessary minimum.
The only drawback is images could become a little larger, require some more memory and CPU to be edited - usually not an issue given the size and power of actual systems. Remember, anyway, that while monitors capable to display the full AdobeRGB space do exist, no monitor is able to display ProPhoto RGB - while there are ways to display areas that are falling outside a color space boundaries, it could become impossible to show subtle differences.