insurance-weekly-report
Pass
Audited by Gen Agent Trust Hub on Feb 20, 2026
Risk Level: SAFE
Full Analysis
- [Indirect Prompt Injection] (SAFE): The skill identifies multiple ingestion points for untrusted data, including Excel, CSV, JSON, and DuckDB files. While this presents an injection surface, the skill's primary function is structured data aggregation and PowerPoint generation, which minimizes the risk of the agent executing embedded instructions. Evidence: Ingestion points: User-uploaded data files (README.txt); Boundary markers: Not explicitly defined in config; Capability inventory: PPT generation (generate_report.py) and data transformation (data_transformer.py); Sanitization: Not specified in these configuration files.
- [Dynamic Execution] (SAFE): The
field_mappings.jsonfile defines mathematical formulas for KPI calculation, which indicates the use of a dynamic evaluation engine. However, these formulas are sourced from the skill's own static, trusted configuration rather than user input, mitigating potential code injection risks. Evidence: Formula definitions in field_mappings.json; Formula parser improvements noted in CHANGELOG.md.
Audit Metadata