Anything I am missing?
My only concern would be that the deck be truly and well shuffled.
Cards sticking together, shuffling habits, and other human factors could influence the realistic number of permutations available, and might reduce the total entropy of the seed.
The question (perhaps one that can only be answered by running hundreds of thousands, or millions, of tests with actual humans shuffling actual decks of cards) is: How much entropy is lost to the various biases that might be involved?
Randomness is a good point. If you use a perfect shuffle, it isn't random of course. But if you model a random riffle shuffle mathematically seven shuffles should be sufficient (cut binomially, drop cards from either side with a probability proportional to the current size of each side). I remember seeing the proof that at that point any combination is equally possible. Of course no one would be that good so another couple of shuffles should be done when doing it by hand.