The Agent Skills Directory

[SAFE]: The skill functions as a best-practice guide for pandas development, focusing on vectorization and memory efficiency. All code samples provide standard, safe implementations of data manipulation tasks.- [NO_CODE]: This skill consists of markdown documentation and code templates. It does not provide automated scripts or executables that run without user intervention.- [DATA_EXFILTRATION]: No network operations or attempts to access sensitive system files (e.g., credentials, SSH keys, or environment files) were found. Data operations are localized to the user's provided DataFrames.- [INDIRECT_PROMPT_INJECTION]: The skill describes patterns for ingesting external data, which is a common attack surface.
Ingestion points: Methods like pd.read_csv and pd.read_parquet are documented in references/performance-optimization.md and references/data-cleaning.md.
Boundary markers: The skill explicitly instructs users to "Validate data quality before and after transformations" and "Check for unexpected groups."
Capability inventory: Capabilities are limited to pandas-based file I/O and data transformations.
Sanitization: The skill provides an extensive reference for data cleaning (references/data-cleaning.md) and recommends schema validation using the pandera library to ensure data integrity.- [DYNAMIC_EXECUTION]: The performance guide mentions pd.eval() and df.eval() for accelerating calculations on large datasets. These are standard pandas features and are presented strictly in the context of mathematical expression optimization on local DataFrames.

pandas-pro