data-comparability
Pass
Audited by Gen Agent Trust Hub on Aug 16, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external tabular data (CSV, Parquet, Excel, JSON) to analyze headers, types, and categorical values. While this creates a surface for indirect prompt injection, the instructions limit the agent's scope to structural analysis rather than interpreting text as instructions. 1. Ingestion points: Path-based access to user datasets. 2. Boundary markers: Not specified in the procedure. 3. Capability inventory: File system read access for data and write access for report generation. 4. Sanitization: Not explicitly defined for column content.
- [UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill specifies dependencies on pandas and pyarrow, which are standard, well-maintained libraries from the official Python package registry.
Audit Metadata