I like to make all my origin, sizing, and polygon collisions set up to my first animation frame.
Then I copy the frame or animation and make edits to the image. If my next frame image is smaller I dont resize the image, and usually polygon and origin stay the same through out (exceptions do exist).
If the image is larger (ie. character puts their hands up above their head) I will increase the image size and verify the character does not change by previewing before I go on (if origin is center and you stretch the image while having align center it shouldn't change).
It's all about finding a process that works for you ^^