---
name: auto-documentation
description: Automatic documentation generation — data dictionaries, dataset READMEs, transform docstrings, and pipeline changelogs. Use at the end of any pipeline to leave full project traceability. Produces repository/wiki-ready Markdown.
---

# Auto Documentation

Leave every finished pipeline fully documented and traceable.

## Deliverables
- **Data dictionary** (`docs/data_dictionary.md`) — one row per column: name, dtype, missing %, example values, and a business description when supplied.
- **Pipeline changelog** (`docs/pipeline_changelog.md`) — one entry per transformation step with `rows_before` / `rows_after`, generated from the lineage log.

## Rules
- Always run at pipeline completion to preserve traceability.
- Pass a `descriptions` dict for business-critical columns so the dictionary is meaningful.
- The changelog must capture `rows_before`/`rows_after` for every step (sourced from the **data-governance** lineage log).

## Reference implementation
`skills/auto_documentation/documentation.py` (`generate_data_dictionary`, `generate_pipeline_changelog`).