A raw screenshot has hard edges that run straight into the page background and no depth at all, so it reads as a foreign object rather than an illustration. Four small things fix it: padding, a rounded corner, a soft shadow, and a backdrop of its own.
That is exactly what this does, and it does it to the layer or the whole document you are already working on rather than to a file you have to upload somewhere. Frames are optional: none, browser, window, phone or a decorative one, with a dark chrome switch for dark interfaces.
The backdrop is the part people underestimate. It can be a solid color, transparent, or a generated mesh gradient built from the colors in the screenshot itself, which is why the results look considered rather than pasted onto a random purple.
Everything opens in a new document with the layers intact, so annotating afterwards is normal editing: draw an arrow, add a callout, blur the customer name. Nothing is baked.