mechanical-preprocess

What problem does it solve?

This skill handles mechanical cleanup of manuscript text at scale, addressing em-dash usage, forbidden words, and repetitive structures before AI agents touch the text. It uses a bash-first approach to perform safe, deterministic transformations while preserving voice and enabling later AI review.

Core Features & Use Cases

• Bash-first cleanup: pattern matching, counting, and safe replacements across chapters.
• Phase-based workflow: scan, replace, AI-quality-review, and reporting with diffs.
• Output & traceability: generates diffs per chapter and a final mechanical-preprocess-report to guide manual review.
• Use Case: Run this before dialogue polishing to reduce mechanical noise across 30+ chapters.

Quick Start

Run the mechanical-preprocess workflow on your chapters directory to clean em-dashes, forbidden words, and repetition before AI editing.