Code Documenter finds undocumented code, asks which docstring format you want, and applies it consistently. It covers Google, NumPy and Sphinx docstrings for Python, JSDoc for TypeScript, and API documentation for FastAPI, Django, NestJS and Express.
Examples in the documentation are validated: doctest for Python, tsc for TypeScript and the Redocly linter for OpenAPI. The run ends with a coverage summary.
What You Get
- Documented files with parameter types, return values and raised errors
- OpenAPI specs and developer portal configuration
- Doc site setup with content structure and build instructions
- Getting started guides, tutorials and troubleshooting pages
- A documentation coverage report
Common Use Cases
- Adding Google-style docstrings to a Python module
- Writing JSDoc for a TypeScript library
- Producing API docs for a FastAPI or NestJS service
- Building a documentation site with Docusaurus or MkDocs
- Writing a getting started guide
What Makes It Unique
It treats documentation as something to test. Python examples run under doctest or pytest, TypeScript examples compile with tsc, and OpenAPI specs go through a linter. Eight reference files split the work by language, framework, coverage reports, doc systems, interactive API docs and guides.
Under The Hood
The skill is a SKILL.md with eight reference files and no scripts of its own. During a run the agent edits your source files to add documentation and runs validation commands in your shell: python doctest, pytest, tsc and npx for the Redocly linter, which can download that npm package. It does not fetch web pages or call outside APIs.