artifact-clarity-pass — cut AI slop down to real signal

Take verbose, generic, hedged, over-explained output and tighten it to what actually carries meaning. The job is subtraction with judgment: remove the filler, keep every load-bearing word, fact, edge case, and caveat.

⚠️ Critical Constraints

• Never cut meaning to cut words. Slop is text that carries no information. A real caveat, edge case, or constraint is signal — keep it even if it's long. Why: the failure mode of de-slopification is deleting a load-bearing "but only if X" along with the fluff around it. Removing words is easy; knowing which words mattered is the whole skill.
  • WRONG: cut "Note that this fails silently if the file is missing" (lost a real edge case)
  • CORRECT: cut "It's worth noting that, generally speaking, in most cases" (pure filler)