data-exploration-visualization
Warn
Audited by Gen Agent Trust Hub on Mar 12, 2026
Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill uses
joblib.load()inscripts/modeling_evaluator.pyandscripts/data_preprocessor.pyto load serialized machine learning models and preprocessing pipelines. Sincejoblibrelies onpickle, loading untrusted files can lead to arbitrary code execution. - [COMMAND_EXECUTION]: The unsafe deserialization mechanism in
joblib.load()allows for arbitrary command execution on the host system if an attacker provides a maliciously crafted model file. - [PROMPT_INJECTION]: The skill has a vulnerability surface for indirect prompt injection (Category 8).
- Ingestion points: Data enters the system through
pd.read_csv,pd.read_excel, andpd.read_jsoninscripts/eda_analyzer.py. - Boundary markers: The system lacks delimiters or instructions to prevent the agent from following commands embedded in the data.
- Capability inventory: The skill can perform file writes via
scripts/report_generator.pyand execute modeling logic inscripts/modeling_evaluator.py. - Sanitization: There is no evidence of content sanitization or validation for processed data strings before they are used in report generation.
Audit Metadata