Your schema.yml describes what your models do. But why you chose cash over accrual accounting, what that hardcoded filter is working around, why that staging model exists at all — that’s in Slack threads, someone’s head, or nowhere.
dbt Zettelkasten is a plug-and-play skill for AI coding agents like OpenCode and Claude Code. Drop it into your agent’s skills directory and it can autonomously document your pipeline as a web of interlinked Markdown notes. For every model, the agent reads the raw SQL and generates three linked notes: the technical lineage, the business logic, and the tech debt.
It also ships with a query skill — ask your agent real questions like “Trace the origin of opportunity_rsac” or “What breaks if I switch revenue recognition to accrual?” — and get answers grounded in your actual codebase, not hallucinations.
Models get refactored? Deprecated notes get tombstoned with bidirectional links, so you never lose historical context.
If you’re running OpenCode or Claude Code for dbt development and want domain knowledge to live in the codebase instead of outside it, give it a look: [LINK]
Feedback and contributions welcome!