We could do a HEAD request to all image URL's to ensure the mimeType returned is actually an image.
This would prevent things like #430 where the URL added didn't actually point to an image 👍
Existing tests: https://github.com/electron/electron-apps/blob/master/test/human-data.js#L100