The Agent Skills Directory

[DYNAMIC_EXECUTION]: The scripts/pipeline_orchestrator.py script generates Python code for Airflow, Prefect, and Dagster workflows based on user-supplied configurations. It utilizes the compile() function to validate the syntax of the generated code strings at runtime. This practice of compiling dynamically generated code from variable inputs represents a security risk if those inputs (such as table names or connection identifiers) are manipulated to include malicious code.
[INDIRECT_PROMPT_INJECTION]: The skill provides tools that ingest data from external sources, which could be used to deliver malicious instructions to the agent.
Ingestion points: scripts/data_quality_validator.py (loads CSV, JSON, and JSONL data), scripts/etl_performance_optimizer.py (loads SQL queries and JSON metrics), and scripts/pipeline_orchestrator.py (loads YAML and JSON configurations).
Boundary markers: The scripts use standard format parsers but do not implement explicit boundary markers or "ignore" instructions to prevent the agent from being influenced by content within the data files.
Capability inventory: The skill has the capability to write files (generated code and analysis reports) and execute complex data processing logic.
Sanitization: There is no specific evidence of sanitization or filtering logic designed to strip instructions or malicious prompts from the data files before processing.

senior-data-engineer