The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to install the polars library and its optional dependencies (e.g., excel, database, fsspec, pandas, numpy) using the uv package manager from standard Python registries. It also recommends fetching citation metadata from the official arxiv.org repository.
[INDIRECT_PROMPT_INJECTION]: The skill enables the ingestion of external data from multiple formats, including CSV, JSON, Parquet, and Excel, which serves as a potential attack surface for indirect prompt injection if the data contains malicious instructions for an agent.
Ingestion points: read_csv, scan_csv, read_parquet, scan_parquet, read_json, read_ndjson, scan_ndjson, read_excel, read_database, and read_database_uri (found in references/io_guide.md).
Boundary markers: None explicitly defined for natural language instructions embedded within data fields.
Capability inventory: Supports local file writes, cloud storage uploads (AWS S3, Azure Blob, GCS), database updates, and SQL execution across multiple scripts.
Sanitization: The skill emphasizes data integrity through strict schema validation and type casting (e.g., schema_overrides), though it does not explicitly address the sanitization of natural language content for downstream AI processing.
[COMMAND_EXECUTION]: The skill provides numerous patterns for data transformation and SQL query execution using the Polars library's native expression API and SQLContext.

polars