The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill connects to cartblanche22.docking.org and files.docking.org to retrieve chemical compound data and 3D structures. These are official domains for the ZINC database, a well-known academic resource maintained by the University of California, San Francisco (UCSF).
[COMMAND_EXECUTION]: The skill utilizes a bundled Python script scripts/query_zinc.py to handle API interactions. The script uses only standard Python libraries (urllib) to perform form-encoded POST requests and does not contain any risky execution patterns.
[INDIRECT_PROMPT_INJECTION]: The skill processes data (SMILES strings, molecular properties, and ZINC IDs) fetched from an external API. While this technically constitutes an indirect prompt injection surface should the external source be compromised, the risk is negligible given the structured nature of the data and the scientific context.
Ingestion points: API responses processed in scripts/query_zinc.py and documentation examples using pandas to read TSV files.
Boundary markers: None explicitly defined for chemical strings, but data is typically parsed into structured JSON/DataFrames.
Capability inventory: Limited to network retrieval and local script execution for data processing.
Sanitization: Standard URL encoding is used for outbound API parameters.

alterlab-zinc-db