litho-document-skill

Invoke this Skill with a request to document the repository. The agent scans the project structure, researches system context and domain modules, then writes C4 context, container, and component diagrams plus workflow and interface documents as Markdown files.

It produces six outputs: project overview, architecture document, workflow document, a Deep-Exploration folder with one file per domain module, boundary interfaces, and a conditional database overview with ER diagrams when SQL files or ORM dependencies are detected.

No. It is a pure agent-side parallel implementation of deepwiki-rs that relies entirely on the agent's own tools such as file reading, grep search, and semantic codebase search, with no external binary dependency.

Yes. It adapts scanning strategy by project size: full reads for small projects, outline views and targeted searches for large ones. Research findings are persisted to a .litho-agent/ directory so long sessions do not lose earlier analysis.

It is generated when the project contains .sql files, migrations or db directories, ORM dependencies like sqlx or prisma, or database configuration files. Otherwise a minimal declaration file is written stating no database was detected.

Analysis quality depends on how much code the agent actually reads, so low-importance modules receive only brief coverage. Each document includes a confidence score from 1 to 10 and a list of items needing human review.