Image manipulation is a good thing, but it's useless if the effect is only temporary. We want to be able to manipulate the image for uses such as placing in a web page, using in a gallery, or as a profile image.
Again, there are multiple ways to do this. We could manipulate the original image, saving changes back to the image, but that's destructive. By repeatedly manipulating the same file, you risk (actually, it's much more certain than risk) losing information. For example, resizing something down, and then up, results in pixelated or blurred images. And cropping—well, once something is cropped, you can't recover the uncropped image.
The solution I'd recommend is to leave the original alone, and create a new image with the effects applied to it. This allows you to make multiple different versions of the same file, and also allows you to fine-tune the effects if they were a bit off the first time you did them.
To do this, I would recommend creating a second repository...