Suppose you’re implementing a “Wordle-like” game such as Juho Snellman’s Huewords or Alan Bellows’ Omiword. The defining principle of such games (as I’m defining them here, anyway!) is that the solution to the puzzle is an arrangement of English words which satisfies some criterion, and the player arrives at that solution by guessing various English words, such that the game must differentiate “that’s a valid word, but not the answer” from “that sequence of letters isn’t even a word.”

All of Wordle, Huewords, and Omiword take the additional step of programmatically generating new puzzles based on a randomized algorithm. (Wordle is the simplest in all respects: its solution is a single word, and its puzzle-generation algorithm is “pick a random word from this list of target words.”)

If you’re programming this kind of game, you absolutely must maintain two word lists: a list of valid words that may appear in the player’s guesses, and a smaller list of target words that may appear in the solution to a programmatically generated puzzle. The latter list will be a subset of the former: just those words that are both “fair” and “appropriate” to appear in an intended solution. By “fair” I mean that we don’t want the answer to the daily Wordle to be, say, ETUIS. Nobody knows that word; it’s not fair! By “appropriate” I mean that we don’t want the answer to be… well, fill in your favorite inappropriate five-letter word here. In crossword construction, we speak of the breakfast test (sometimes the “Sunday morning breakfast” test), which means you shouldn’t put anything in a crossword grid you wouldn’t want the proverbial little old lady to encounter over the breakfast table.

The breakfast-test criterion should be pretty stringent, and in crossword construction pertains to not just the clue but also its entry. Thus, a crossword shouldn’t contain, say, COCKS, even if it’s clued as “Roosters.” The other example I would have used here was “ABO, even if it’s clued as ‘Common blood types’ ”; but actually that’s quite common in the NYT these days, even if it took a relative hiatus in the 1990s and hasn’t been clued as “Australian native” since 1980.

However, such “inappropriate” words should absolutely remain on your list of valid words! When in doubt, put the word on your valid-word list; and when in doubt, take it off your target-word list. This avoids the following failure modes:

• Suppose you have a questionable word (say, COCKS) on your target-word list. Then it might appear in your puzzle’s intended solution. But the player, if he’s familiar with crosswording conventions, might reasonably assume that the intended solution can’t possibly include COCKS — so, he won’t pursue that direction, and might frustratedly conclude that the puzzle is unsolvable. In other words, correctly solving the puzzle should not require entering profanity.

• Suppose you have a questionable word absent from both lists. Then when the player enters the word, your game will tell him it’s “not a valid word,” which (depending on your player’s level of dirty-mindedness) might just seem like a bug. For example, Huewords’ valid word list originally lacked the word TRAMP — to Huewords’ author primarily a breakfast-test-failing synonym for “skank,” but to this player merely an innocuous word for “hobo” blessed by Sinatra and Disney.

• Suppose you have a questionable word absent from both lists. When you generate a new puzzle using words from the target word list, naturally you’ll ensure its solution is unique: there must be no arrangement of words from the valid word list that produces an equally acceptable solution. (This doesn’t apply to Wordle, but does to Huewords and Omiword.) Now, if it’s missing from your valid word list, it’s possible that there will be an alternative solution involving the word SKANK! This will at least be a source of amusement to the player. In other words, any word missing from the valid word list may appear in a “cook” of the puzzle, which is just as interesting as the intended solution — perhaps even more so.

Essentially, your game should be liberal in permitting the player to enter profanity, but profanity should never help the player to find either a solution or a “cook.” This puts everyone’s incentives in the right places. And the way you do this is by having a restrictive target-word list plus a liberal valid-word list.