Not entirely sure what you mean in here. Yes, it is enough if we have the first four letters of a word, because they are unique. So, we have more room to introduce information that will obscure these letters? For example, could URU/Yw be "aren" followed by another input that is used in your rule to help us go figure out the next word?

I've been looking at it since yesterday, but it's a loss of time until this point. We know absolutely nothing about your rule. It could be anything, like take the first cipher-letter ('U') and use it to find potential letters that it points given a combination using the silver and black cards (e.g., from the first of the twelve combinations, it points to 'h', 'j', 'n' and 'm'). That's just one of the nearly infinite rules I can think of.

I checked an account under the nostr post you shared, from BitCat, which I presumes is yours, because they're describing a very similar patent of securing passwords. As far as I can tell, it's infeasible to achieve reversal; our only hope is that you've used a "simple" rule, which could be utterly subjective.