mandoline

What problem does it solve?

This Skill automates the transformation of raw, opaque survey data and CSV dumps into self-documenting, clean BigQuery tables, preventing LLM hallucinations and enabling direct analysis.

Core Features & Use Cases

• Schema Design: Guides the creation of descriptive column names, appropriate data types, and decodes numeric codes into human-readable labels.
• Data Cleaning: Handles filtering of junk rows, conversion of placeholder values to NULL, and snapshotting of fragile Sheets-linked data.
• Metadata Enrichment: Writes teaching-style column descriptions, table descriptions, friendly names, and labels for enhanced discoverability and usability.
• Automated Verification: Includes a suite of SQL checks to ensure data integrity and transformation accuracy.
• Use Case: When presented with a raw survey export containing columns like S1, Q1r3, and coded responses, use Mandoline to create a BigQuery table where columns are named gender, action_physical_activity, and values are clearly labeled 'Male', 'Female', 'Yes', 'No', with rich descriptions explaining their meaning and usage.

Quick Start

Use the mandoline skill to transform the raw data in the 'raw_survey_data' table into a clean, analysis-ready table with proper descriptions and metadata.