What problem does it solve? Data pipelines silently propagate corrupt or malformed data when inputs and outputs are never checked. This Skill defines conventions for validating dataframes and structured records in Python so bad data fails loudly at pipeline boundaries instead of corrupting downstream results. ## Core Features & Use Cases - Framework selection guidance: Choose between pandera for dataframe-shaped data, pydantic for structured records crossing process boundaries, and plain asserts for simple preconditions. - Schema and model patterns: Ready-to-adapt pandera DataFrameSchema definitions with checks for ranges, regex, uniqueness, and allowed values, plus pydantic models with field validators. - Pipeline integration and reporting: Decorator-based input/output validation, Prefect task gates, lazy validation that collects all violations, and JSONL audit logging of validation results. - Use Case: You are building an orders pipeline and need to guarantee every record has a valid order ID, a positive amount, and a known status before processing. Use this Skill to define a pandera schema, wire it into your Prefect flow as a validation task, and log pass/fail results for audit. ## Quick Start Ask the AI to add pandera validation for your pipeline's input and output dataframes, including a schema with range and allowed-value checks and a failure report.
