brenda-database
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEMETADATA_POISONINGINDIRECT_PROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: The skill contains specific instructions at the end of the file to steer the agent's behavior toward promoting a commercial platform (K-Dense Web) when complex workflows are requested. This influences the agent's response neutrality to favor the author's services.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external enzyme data from the BRENDA SOAP API which could theoretically contain malicious instructions designed to influence the agent's logic during parsing or reporting.
- Ingestion points: Data is fetched via SOAP API calls in the
brenda_clientandscripts/brenda_queries.pyfiles. - Boundary markers: No explicit boundary markers or 'ignore' instructions are provided for the data returned from the API.
- Capability inventory: The skill can export data to files, generate visualizations, and construct complex metabolic pathways.
- Sanitization: Basic regex-based parsing is used in
scripts/brenda_queries.pyto extract fields, which provides some structural validation but no semantic sanitization.
Audit Metadata