This fixes an edge case in XMonad.Prompt.FuzzyMatch where fuzzySort assumed every candidate matched the input and could fail when that wasn't true. The ranking logic now treats non-matching completions as the worst possible match instead of trying to take the minimum of an empty result set, and the behavior is documented in CHANGES.md. In practice, prompt completion becomes more robust when mixed or partially filtered candidate lists are passed through the fuzzy sorter.