The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/pipeline_orchestrator.py utilizes the Python compile() function within its validation methods for Airflow, Prefect, and Dagster generators. It is used to verify the syntax of the code strings it generates. While used for validation, dynamic compilation of generated code is a sensitive operation that requires careful monitoring of the generation logic.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection (Category 8). Several tools, including data_quality_validator.py, rag_evaluator.py, and dataset_pipeline_builder.py, process external data files in CSV, JSON, and YAML formats. In a production environment, if these files contain malicious natural language instructions and their content is later interpolated into prompts for the agent or other downstream LLMs without strict sanitization or boundary markers, it could lead to unauthorized instruction execution.
[COMMAND_EXECUTION]: The scripts/pipeline_orchestrator.py tool is designed to generate configurations that execute system-level shell commands (e.g., BashOperator) and database queries. This is consistent with its primary purpose as an ETL orchestration tool, but users must manually verify the integrity of the generated DAGs before deployment to ensure no unintended commands are included.

ai-engineering