The Agent Skills Directory

[COMMAND_EXECUTION]: Use of dynamic code compilation. The script scripts/pipeline_orchestrator.py contains calls to the compile() function within the validate methods of the AirflowGenerator, PrefectGenerator, and DagsterGenerator classes. This function is used to check for syntax errors in generated pipeline code. While the code is compiled for validation rather than execution, runtime compilation of generated source code is a sensitive operation.
[PROMPT_INJECTION]: Potential for indirect prompt injection due to processing untrusted data and queries. 1. Ingestion points: Untrusted data enters the agent context through file reading operations in scripts/data_quality_validator.py (CSV, JSON, and JSONL data files), scripts/etl_performance_optimizer.py (SQL query files and data statistics JSON), and scripts/pipeline_orchestrator.py (pipeline configuration files). 2. Boundary markers: Absent across all scripts. 3. Capability inventory: File system writes are performed in scripts/data_quality_validator.py and scripts/etl_performance_optimizer.py. The scripts/pipeline_orchestrator.py script generates executable Python code and performs dynamic compilation using compile(). 4. Sanitization: Absent; the skill relies on standard parsers without specific sanitization for embedded natural language instructions.

senior-data-engineer