What problem does it solve? Python projects often accumulate inconsistent data I/O code: hardcoded paths, dtype guessing on CSV reads, unparameterized SQL, and credentials scattered through source files. This Skill establishes a single set of conventions for file-based I/O, database connections, path handling, and credential management so every data read/write follows the same safe patterns. ## Core Features & Use Cases - Format selection guidance: Decision table for choosing parquet (polars), CSV, Excel, JSON, or pickle based on data shape and audience, with parquet as the default for tabular data. - Database connectivity: Parameterized SQLAlchemy queries with connection pooling, plus DuckDB for querying parquet files directly without loading them into memory. - Safe path and credential handling: Enforces pathlib.Path over string concatenation and environment-variable credentials loaded via python-dotenv, never hardcoded secrets. - Use Case: You need to read a folder of partitioned parquet files, filter to one store, and write the result to PostgreSQL. The Skill provides the lazy polars scan with predicate pushdown, the SQLAlchemy engine pattern, and the write_database call in one consistent style. ## Quick Start Ask the AI to write Python code that reads a parquet file with polars, filters it, and writes the result to a database using environment-variable credentials.
