model-explanation

Pass

Audited by Gen Agent Trust Hub on Jun 29, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill performs model interpretability analysis using standard libraries like SHAP and XGBoost to generate reports. It does not perform network operations or unauthorized file access.\n- [DYNAMIC_EXECUTION]: The script modifies sys.path to include the _vendor directory for internal dependency resolution in explainer.py. This is restricted to the skill's local directory structure.\n- [DYNAMIC_EXECUTION]: The skill uses pandas.query() in explainer.py to filter datasets based on CLI arguments. This is standard functionality for data analysis and is scoped to the loaded data.\n- [INDIRECT_PROMPT_INJECTION]: The skill possesses an ingestion surface for untrusted data.\n
  • Ingestion points: model_path, data_path (explainer.py)\n
  • Boundary markers: Absent\n
  • Capability inventory: File read, Matplotlib image generation, Markdown/JSON file write (explainer.py)\n
  • Sanitization: None (data loaded directly into Pandas and SHAP)
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 29, 2026, 06:41 AM
Security Audit — agent-trust-hub — model-explanation