Generate dbt Validation SQL Notebooks

This skill generates a Monte Carlo SQL Notebook containing validation queries for dbt model or snapshot changes, sourced from either a GitHub PR URL or a local dbt repository (auto-detected from the target argument), and outputs it as a browser-openable import URL (<MC_BASE_URL>/notebooks/import#<base64-encoded-yaml>). It requires the GitHub CLI (gh, authenticated) for PR mode, Python 3 with pyyaml, and works best invoked with /model sonnet for faster generation.

The workflow runs in six phases: Phase 1 fetches changed .sql files under models//snapshots/ (via gh pr view/gh pr diff for PR mode, or git diff against the detected base branch for local mode), plus dbt_project.yml for schema resolution, filtered to at most 10 models unless the caller specifies a --models list; Phase 2 parses each changed model's metadata, diff, and classifies it as NEW or MODIFIED; Phase 3 generates validation queries per classification - NEW models get row-count, sample-data-preview, segmentation-count, uniqueness-check, NULL-rate, and time-axis-continuity queries, while MODIFIED models additionally get changed-field-distribution and before/after comparison queries between {{prod_db}} and {{dev_db}}; Phase 4 assembles the notebook YAML (metadata, parameter cells for prod_db/dev_db database selection, a markdown summary cell, and SQL cells organized by model); Phase 5 encodes it into an import URL; Phase 6 presents a summary and opens the notebook.

cells:
  - id: param-prod-db
    type: parameter
    content:
      name: prod_db
      config: { type: text, default_value: "ANALYTICS", placeholder: "Prod database" }

Generated SQL uses ANSI-compatible syntax working across Snowflake, BigQuery, Redshift, and Athena, with minor adjustments possibly needed for warehouse-specific quirks. Two helper scripts support the flow: resolve_dbt_schema.py (resolves each model's output schema from dbt_project.yml routing rules and model config overrides) and generate_notebook_url.py (encodes the notebook YAML and opens it in the browser).

Use it to validate dbt model or snapshot changes before or after merging - reviewing a PR's changed models, or comparing dev vs. prod state for a local branch's changes - by generating targeted before/after and data-quality queries rather than writing them by hand. It is scoped to changed .sql files under models//snapshots/ directories; non-model changes are not analyzed, and if no model SQL files changed, the skill reports that and stops rather than generating an empty or irrelevant notebook.

Input is a GitHub PR URL or a local dbt repo path, with optional --mc-base-url and --models filters. Output is a Monte Carlo SQL Notebook import URL containing parameterized validation queries (row counts, uniqueness, NULL rates, segmentation, and before/after comparisons) for each changed model, ready to open and run interactively in the MC Bridge SQL Notebook interface.

Uses the GitHub CLI (gh) for PR metadata/diffs/file content, git for local-mode diffs, Python/pyyaml for notebook YAML construction, and targets Monte Carlo's SQL Notebook interface with ANSI-compatible SQL portable across Snowflake, BigQuery, Redshift, and Athena.

Data/analytics engineers reviewing dbt model changes (in a PR or locally) who want an automatically generated, ready-to-run set of validation queries comparing before/after state rather than writing ad hoc SQL checks by hand.