Анализ исходников для генерации README

documentation

readme-analyze-source is a Claude Code skill that scans a module's source directory and produces five structured artifacts: a spine candidate (the mental-model paragraph for the README), a list of callout candidates with file-and-line references, an API surface inventory, section clusters grouped by use rather than API shape, and running-domain seeds. It takes a source path via `<source-dir>` and deliberately skips any existing README to avoid anchoring the spine on prior wording. The output is consumed directly by `readme-draft` and `readme-analyze-existing`; the skill itself neither drafts documentation nor evaluates what already exists. Source data comes from scaladoc on package objects and public types, load-bearing method signatures, test-case names that describe edge behavior, and implementation comments.