tooluniverse-metabolomics
Pass
Audited by Gen Agent Trust Hub on Feb 20, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- PROMPT_INJECTION (LOW): Identified surface for indirect prompt injection via ingestion of external database content.
- Ingestion points: The skill retrieves metabolite and study information from external sources including HMDB, MetaboLights, and PubChem (referenced in
test_tools_debug.pyandtest_detailed.py). - Boundary markers: Sample reports like
diabetes_metabolomics_report_FIXED.mddemonstrate that external data is directly interpolated into Markdown without protective delimiters or instructions for the agent to ignore nested commands. - Capability inventory: The skill demonstrates the capability to write research reports to the local file system (using the
output_fileargument in the pipeline defined indiabetes_analysis.py). - Sanitization: There is no evidence of sanitization, escaping, or validation of the text retrieved from external APIs before it is included in the final research reports.
Audit Metadata