senior-data-scientist
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The
SKILL.mdfile contains standard shell commands for model training, testing, and deployment (e.g.,docker,kubectl,helm,pytest). These are consistent with the skill's stated purpose of production-grade data science and MLOps. - [INDIRECT_PROMPT_INJECTION]: The skill scripts (
scripts/experiment_designer.py,scripts/feature_engineering_pipeline.py,scripts/model_evaluation_suite.py) define input parameters that process external data. While the current implementation consists of boilerplate code with no immediate risk, this creates a standard surface for processing potentially untrusted data. - Ingestion points: Files provided via the
--inputargument in the provided Python scripts. - Boundary markers: None explicitly defined in the boilerplate code.
- Capability inventory: Local file reading and writing, standard logging.
- Sanitization: Standard Python exception handling is implemented, though specific data validation logic is currently placeholder code.
Audit Metadata