What problem does it solve? Long-form Chinese manuscripts often accumulate AI-flavored clichés, absolute commands, mechanical punctuation, and cross-chapter contradictions that are hard to catch manually. This Skill provides a script-backed quality audit methodology so revision decisions rest on measurable evidence instead of subjective claims. ## Core Features & Use Cases - Four-layer quality scoring: Evaluates sentence (S1-S6), paragraph (P1-P4), chapter (C1-C4), and whole-book (B0-B5) rules with a 7.5 pass threshold via make quality. - Cross-chapter and whole-book consistency: Checks CX-1 to CX-7 for conflicting claims, duplicated objects, and terminology drift, plus a whole-book ledger covering dash density, terms, claims, and release alignment. - De-AI-flavor cleanup: Uses machine-readable lexicons (directive-lexicon.json, cliche-catalog.json) to locate buzzwords, absolute imperatives, and template phrasing for human review. - Use Case: After drafting an eight-chapter Chinese book, run the audit to find chapters with excessive em-dash density, repeated case studies across chapters, and unresolved TODO placeholders before delivery. ## Quick Start Ask the AI to run a quality audit on the manuscript directory and report S/P/C/B scores, gate status, and cross-chapter consistency issues with evidence locations.
