What problem does it solve? LLM-drafted prose is structurally bloated before it is wordy, so sentence-level trimming lands at 80% of the original length and still reads padded. This Skill rebuilds a document from an explicit inventory of its claims, reasons, instructions, and evidence, then accounts for everything it dropped so no substance is silently lost. ## Core Features & Use Cases - Claim inventory and classification: Extracts every claim, reason, instruction, and piece of evidence, then classifies the remainder into six bloat kinds (padding, restatement, abstraction inflation, hedge sprawl, structural bloat, false parallelism) each with its own fix. - Rebuild with drop accounting: Rewrites the document from the inventory rather than trimming the draft, then reports every dropped item with its reason so nothing disappears silently. - Protected content rules: Never cuts reasons attached to rules, exact enumerations, numbers with the commands that produced them, named alternatives, or stated uncertainty. - Use Case: A 94-word paragraph about error handling distills to 38 words carrying the same four claims, with the dropped synonym doublets reported and the missing rationale flagged as a finding. ## Quick Start Distill this draft design doc down to what it actually claims and tell me everything you dropped.
