What problem does it solve? Long-form non-fiction writing projects (reports, manuals, whitepapers, books) often lose track of progress, produce AI-sounding filler prose, or break consistency across chapters. This Skill enforces a disk-based writing loop where outlines, chapter ledgers, and manuscript files are the source of truth, and every batch passes structural and style checks before being called done. ## Core Features & Use Cases - Structured Writing Loop: Reads outline.md, chapter-ledger.md, and existing manuscript files from disk before each batch, writes one to two chapters per cycle, and verifies results with make write. - Continuation, Expansion, and Refinement: Continues chapters from disk state rather than chat memory, expands content via a gated amplification plan with measured character counts, and refines drafts with backup-first edit protocols. - Genre Guides and Style Constraints: Loads genre-specific references (whitepaper, training, report, memoir, genealogy, consultant, ghostwriter) and enforces Chinese heading hierarchy, anti-AI-pattern prose rules, and table/bold usage limits. - Use Case: A user asks to continue writing chapter 3 of a market research whitepaper. The Skill reads the outline and ledger from disk, writes the chapter into manuscript/ch03-*.md, runs make write to validate headings, word count, and placeholders, then updates the ledger. ## Quick Start Continue writing the next chapter of my whitepaper project in the current workspace, following the outline and chapter ledger on disk.
