zinc-database

Pass

Audited by Gen Agent Trust Hub on Jun 26, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides instructions and Python templates to execute shell commands such as curl, wget, and aria2c. These are used specifically to interact with the CartBlanche22 API and ZINC file repositories for retrieving chemical compound data and 3D molecular structures.
  • [EXTERNAL_DOWNLOADS]: The skill facilitates downloading molecular structures and compound information from established domains including cartblanche22.docking.org and files.docking.org. These domains are official resources for the ZINC database maintained by the UCSF Department of Pharmaceutical Chemistry.
  • [PROMPT_INJECTION]: The skill processes external data from the ZINC API which presents a surface for indirect prompt injection.
  • Ingestion points: Data fetched from cartblanche22.docking.org is loaded into the agent's context and parsed via TSV loaders in SKILL.md and api_reference.md.
  • Boundary markers: API responses are processed as structured TSV data, though no explicit 'ignore embedded instructions' markers are used.
  • Capability inventory: The skill includes the ability to perform further network requests and execute local subprocesses using curl and wget.
  • Sanitization: The provided Python integration examples use subprocess.run with list arguments, which prevents shell command injection from malformed SMILES strings or identifiers.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 26, 2026, 07:59 PM