The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/pipeline_orchestrator.py uses the compile() function to validate the syntax of generated pipeline scripts for Airflow, Prefect, and Dagster. This is a common pattern for code generation tools to ensure the output is syntactically correct and does not involve the execution of the compiled code within the analyzer's environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data for analysis and configuration generation, which creates a potential surface for indirect prompt injection.
Ingestion points: The scripts scripts/data_quality_validator.py and scripts/etl_performance_optimizer.py read data from local CSV, JSON, and SQL files, as well as Spark job history logs.
Boundary markers: No explicit boundary markers or instructions to ignore embedded prompts are used when reading data files.
Capability inventory: The skill can generate configuration files and perform statistical data analysis, but it does not directly execute code derived from these inputs on the host system.
Sanitization: Data is parsed using standard Python libraries (json, csv) or regex-based SQL parsing. While this mitigates basic injection attacks, it does not prevent malicious content in the data from potentially influencing the logic of the generated analysis reports or pipeline configurations.

senior-data-engineer