Categories and tags accumulate cruft. Terms that once had posts and now have none, near-duplicates that crept in over years of different people filing content, a plural sitting next to its singular. None of it is urgent, so none of it ever gets cleaned, and the taxonomy list slowly turns into something nobody wants to open.
This is the read-only survey that makes the cleanup possible. It counts every term, surfaces the empty ones and the likely duplicates, and hands you a short list to act on. It cannot change anything.
Audit the [category] taxonomy for me, read only, change nothing. List everyterm with its post count, then call out two things: the empty terms with acount of zero that are just cluttering the list, and the likely duplicates,terms whose names or slugs are near-matches of each other, like a singular anda plural, or the same word with different casing or a stray typo. For any pairyou are unsure about, read both terms and show me their details side by side soI can decide. Put a short summary at the top: how many terms, how many empty,and how many possible duplicates.Why this is safe to run
Both abilities behind this prompt are reads. Listing the terms with their counts and reading a single term to compare a pair both write nothing, so the survey cannot merge, rename, or delete a thing. That is the point: it is the look-before-you-touch step, and you stay the one who decides what to do with what it finds.
It also does not overreach on the judgment call. It flags duplicates as candidates by comparing names and slugs for near-matches, and where it is unsure it puts the two terms side by side and asks you rather than assuming they are the same. When you are ready to act on the list, the actual cleanup, renaming, merging, or removing a term, is a separate step you take deliberately.
Already connected? Skip the paste. Ask your agent for the “audit a taxonomy for empty or duplicate terms” recipe, or just name the taxonomy you want surveyed, and it runs the same reads on your own install.