Генерация черновика README по анализу кода

documentation

readme-draft is a Claude Code skill that produces a README draft from the structured output of `readme-analyze-source`, strictly limiting itself to what the provided inventory contains and never inventing capabilities or surfaces. The input consists of five sections: a spine candidate, callout candidates, an API surface inventory, section clusters, and running-domain seeds. From these, the skill assembles a curriculum-shaped document — spine prose opens the file, clusters become `##` headings (not individual APIs), and one running-domain value threads through all worked examples. Document length scales with API surface area, from roughly 200 lines for 5–10 inventory entries up to 1500 lines for 40+ entries with cross-platform and integration surfaces. It is intended for new READMEs or full rewrites when an existing file should be discarded entirely.