The Agent Skills Directory

[SAFE]: The skill provides educational content, reference architectures, and boilerplate code for legitimate data engineering tasks. The instructions follow industry best practices for data modeling and orchestration.
[EXTERNAL_DOWNLOADS]: The skill recommends installing standard, widely-used open-source packages such as polars, dbt-core, dbt-snowflake, and apache-airflow from official package registries. These are recognized as safe, well-known tools in the data engineering ecosystem.
[INDIRECT_PROMPT_INJECTION]: The skill demonstrates patterns for ingesting data from external files (CSV, Parquet) and cloud storage (S3 buckets). Ingestion points are identified in examples/python/airflow-data-pipeline.py and scripts/benchmark_dataframes.py. However, since these patterns are used for typical data processing workflows and do not involve the agent executing instructions derived from the data, the risk is negligible.
[COMMAND_EXECUTION]: The provided scripts (scripts/benchmark_dataframes.py and scripts/generate_dbt_models.py) perform local file operations for benchmarking and boilerplate generation. No arbitrary command execution or shell injection vectors were detected.

transforming-data