multiclass-classification

Pass

Audited by Gen Agent Trust Hub on Apr 18, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill follows standard data science workflows using reputable libraries for tabular data analysis and visualization.
  • [EXTERNAL_DOWNLOADS]: The skill references standard, well-known Python packages (scikit-learn, XGBoost, SHAP, etc.) available on official registries.
  • [DATA_EXFILTRATION]: No network operations, sensitive file access, or credential harvesting patterns were found. Data processing is local to the agent environment.
  • [PROMPT_INJECTION]: The instructions focus on technical implementation and do not contain patterns attempting to bypass safety filters or override agent constraints.
  • [INDIRECT_PROMPT_INJECTION]: While the skill processes external tabular data (CSV/Parquet), it does not use dangerous capabilities like shell execution or network requests on the ingested content. Risk is minimal and consistent with standard data analysis tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 18, 2026, 06:43 AM
Security Audit — agent-trust-hub — multiclass-classification