version-dataset
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external datasets (CSV, TSV, Parquet, Excel, etc.) to generate manifests and drift reports. This creates an ingestion surface for untrusted data that could potentially contain malicious instructions aimed at influencing the agent when it reviews the verification output.
- Ingestion points: The
scripts/version_dataset.pyscript reads various tabular data formats using thepandaslibrary. - Boundary markers: The instructions include anti-hallucination rules and specific gates that require the agent to report drift exactly as computed and seek user approval before proceeding.
- Capability inventory: The skill has access to file reading and writing, bash command execution, and text editing tools.
- Sanitization: Content is processed for hashing and logical comparison; the script does not execute or directly interpret the data values as code.
Audit Metadata