Living Docs
Documentation that updates itself when the data changes
Auto-generates and maintains documentation for your data models, SQL queries, and warehouse schemas. When the schema changes, the docs update. When a query is modified, the description stays current. No more stale Confluence pages.
INGREDIENTS
PROMPT
Create a skill called "Living Docs". Connect to my data warehouse and code repository. Generate documentation by: (1) Scanning all tables and columns, generating descriptions from column names, data patterns, and sample values. (2) Analyzing SQL queries and dbt models to extract business logic ("revenue is calculated as sum of order_amount where status = 'completed' and refund_amount is NULL"). (3) Inferring relationships between tables from JOIN patterns and foreign keys to build an ERD. (4) Compiling everything into a searchable data dictionary. On a schedule (or triggered by schema changes), re-scan and update only the sections that changed — preserve any human-edited descriptions. Publish to my configured destination (Confluence, Notion, GitHub Pages, or Markdown files in a repo). Flag columns and tables that have no documentation yet.
How It Works
The problem with data documentation isn't writing it — it's maintaining it.
Docs go stale the moment they're published. This skill generates documentation
directly from your code and schemas, and updates it automatically when things change.
What You Get
- Auto-generated table and column descriptions from schema analysis and usage patterns
- Business logic documentation extracted from SQL queries and dbt models
- An ERD diagram generated from foreign key relationships and JOIN patterns
- A searchable data dictionary
- Automatic updates when schemas change (via Schema Sentinel integration)
- Confluence/Notion/GitHub Pages publishing
Setup Steps
- Ask your Claw to create a "Living Docs" skill with the prompt below
- Connect it to your data warehouse and code repository
- Run the initial documentation generation
- Configure the update schedule and publishing destination
- Optionally integrate with Schema Sentinel for change-triggered updates
Tips
- The auto-generated descriptions aren't perfect — review and edit the initial batch
- Once edited, the skill preserves your human edits and only updates auto-generated sections
- The business logic extraction from SQL is especially valuable for tribal knowledge capture
- Publishing to your team's existing wiki (Confluence, Notion) means people actually find it