1. candidate solution word list, i.e., the entire dictionary used in that version, e.g., OG / GOAT Wordle, the current list for NYT Wordle, whichever variant for which you're solving
2. You're also given an infinite amount of shared solution squares, e.g., as shown in a tweet like this where you can see the gray, yellow, green squares as solved and no letters, to include number of guesses -> https://twitter.com/miradu/status/1500139320472002563
3. can you solve Wordle / prune word list in some meaningful way given that info?
4. specifically, does analyzing the set of blind solutions permit eliminating certain words from the original list -- even one word would suffice
5. you are permitted to leverage any other info that would cryptographically assist, either info can deduce from word list or public domain -- e.g., edit distance from word list, letter frequency tables, etc.
6. you're not given any of the guess words that were used to deduce the blind solutions
7. if the word list can be pruned, by how much -- can half the word list be eliminated in this fashion, none of the word list, etc.
Certainly for some words. If the sought-for word has 3 identical letters (example: “ERROR”), no guess can give you 5 yellow squares (to get 5 yellow squares, you’d need to have 3 R’s in your guess, but there are only 2 incorrect positions for placing an R)
So, if one of the guesses produces 5 yellow squares, the solution isn’t such a word.
(and yes, 5 yellow squares can occur, for example when the word is CARET and the guess TRACE)
Because of the lack of regularity in the word list, I think that fully answering all your questions requires exhaustive analysis. That would be fairly simple, though. For every word in your list, compute the set of blind solutions for all other words, and check which of these sets are unique and which aren’t.