What problem does it solve? Data analysis tasks often require multi-step queries, iterative exploration, and visualizations that are hard to manage in one-off scripts. This Skill provides structured rules for building, executing, and validating Jupyter notebooks, especially when querying BigQuery, so analyses are reproducible, readable, and error-free. ## Core Features & Use Cases - Conditional Execution Workflow: Enforces a generate-cell, execute-cell, validate-output loop when an execution tool is available, or full notebook generation with user-run verification when it is not. - BigQuery Integration Rules: Mandates the %%bqsql BigFrames magic for SQL queries, forbids the Python BigQuery client library, and requires BigFrames (not pandas or scikit-learn) for manipulation and ML. - Notebook Structure & Visualization Standards: Defines a story-like layout with markdown sections, plotting rules, data cleaning constraints, and a mandatory final summary with Q&A, key findings, and next steps. - Use Case: A user asks to explore user retention trends stored in BigQuery. The Skill guides the agent to identify the correct table, query it with %%bqsql, visualize cohort trends cell by cell, and finish with a grounded summary. ## Quick Start Create a Jupyter notebook that analyzes sales trends from my BigQuery table and visualizes the key findings.
