The Agent Skills Directory

[SAFE]: The skill instructions and code snippets focus entirely on legitimate data science workflows using the pandas and numpy libraries.
Snippets provide industry-standard methods for memory optimization, such as chunked reading, categorical type conversion, and downcasting.
Vectorized operations and parquet usage are recommended for performance improvements.
No obfuscation, hardcoded credentials, or unauthorized network activity were found.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process data from external files, which constitutes a standard ingestion surface for this type of tool.
Ingestion points: The skill uses pd.read_csv and pd.read_parquet to load data from files in SKILL.md.
Boundary markers: None are present in the provided code snippets.
Capability inventory: The skill has permissions for Bash and file Write operations.
Sanitization: The snippets do not demonstrate specific sanitization or validation of data content prior to processing.

pandas-coder