What problem does it solve? Writing accurate technical documentation requires deep knowledge of the actual codebase, and AI-generated docs often invent API signatures or carry robotic prose patterns. This Skill grounds every document in real code research and removes residual AI writing patterns before delivery. ## Core Features & Use Cases - Codebase-grounded research: Scans manifests, public APIs, CLI entry points, routes, and existing docs to build a factual research base before any writing begins. - Diátaxis-structured output: Produces tutorials, how-tos, references, explanations, READMEs, or API docs following the Diátaxis framework, with an approval gate before writing starts. - Automatic naturalize pass: Runs a dedicated rewriting agent to strip AI writing patterns while preserving code blocks, tables, and API names. - Use Case: Ask it to "write API reference for the payments module" and it researches the module's exported functions, proposes a doc structure for your approval, writes the reference, and delivers a polished Markdown file under .mz/reports/. ## Quick Start Ask the assistant to write docs for a specific module or polish an existing README, optionally adding a scope or type modifier such as scope:branch or type:tutorial.
