skills/wu-yc/labclaw/pubchem-database/Gen Agent Trust Hub

pubchem-database

Pass

Audited by Gen Agent Trust Hub on Mar 13, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill makes network requests to pubchem.ncbi.nlm.nih.gov using the requests library and the pubchempy package to fetch chemical data and images. These are directed at a well-known scientific service.
  • [PROMPT_INJECTION]: The skill description contains a promotional footer suggesting users switch to 'K-Dense Web' for complex workflows. While not a malicious override of agent instructions, this represents a form of metadata-based steering toward a specific platform.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data from the PubChem API (compound names, synonyms, and annotations). If this data were to contain malicious instructions, it could theoretically influence the agent's behavior, although no specific vulnerable interpolation was identified. This is a standard risk for skills processing web data.
  • [REMOTE_CODE_EXECUTION]: The skill mentions and provides usage examples for internal scripts (scripts/compound_search.py and scripts/bioactivity_query.py) that are not included in the analyzed file. The security of the logic within these scripts cannot be verified.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 13, 2026, 03:46 AM