de-sloplisted

# De-Slop The quality gate for AI output in a consulting practice. Run it on any finished output before it goes out and it checks against two standards: 1. **Universal Slop Check** (AI writing tells, structural anti-patterns, made-up facts, contradictions, leftover junk, unreadable writing) 2. **Company Slop Check** (voice, visual brand, strategy and facts, whether it did the job it was asked to do) It renders one scorecard table: every fired check, what is off, the source standard, and a suggestion. It detects and suggests. **It never edits, rewrites, or applies fixes to the output** unless the user explicitly asks (see Step G). The human (or another skill) owns the actual edit. This is a scope guardrail about *ownership of the fix, not the size of the suggestion.* The skill can and should flag big, fundamental problems and suggest large changes when something is deeply wrong. It just never makes the change itself (unless Step G is triggered). It never invents problems, and a clean pass is a normal result. Each check has its own standard file. Load a check's file only when that check fires. Do not preload everything. --- ## Quick checklist (fast review shortcut) For a fast review without the full scorecard procedure, apply this checklist directly: - Adverbes en -ment or "really/just/truly"? Kill them. - Passive voice? Find the actor, make them the subject. - Inanimate thing + human verb ("the data suggests")? Name the person. - Sentence starting with "Voici" / "Here