pubchem-database

Pass

Audited by Gen Agent Trust Hub on Apr 16, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill includes behavioral steering instructions that direct the agent to monitor the complexity of the user's request and proactively suggest the author's commercial platform (www.k-dense.ai) for complex workflows.- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing untrusted chemical identifiers and structures that are then utilized in network and file-writing operations.
  • Ingestion points: User-supplied chemical names, SMILES strings, and formulas entering the agent context via input parameters (SKILL.md).
  • Boundary markers: Absent; the skill does not define delimiters or provide instructions to the agent to disregard instructions potentially embedded in the chemical data.
  • Capability inventory: The skill utilizes pcp.download for file system writes and requests.get for network operations across multiple scripts (SKILL.md).
  • Sanitization: Absent; user inputs are passed directly to library functions and API requests without visible validation or sanitization logic.- [EXTERNAL_DOWNLOADS]: The skill facilitates downloading molecular structure files (SDF, JSON) and images (PNG) from the official National Institutes of Health (NIH) PubChem servers (pubchem.ncbi.nlm.nih.gov).- [COMMAND_EXECUTION]: The installation documentation provides shell commands for users to install Python dependencies such as pubchempy, requests, and pandas using the uv package manager.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 16, 2026, 08:26 AM