The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides Python code examples that use the subprocess.run method to execute command-line tools like curl, wget, aria2c, and scientific docking software such as dock6 and AutoDock Vina. The implementation uses list-based arguments to pass parameters to these tools, which is a recommended practice to mitigate command injection risks.
[EXTERNAL_DOWNLOADS]: The skill facilitates downloading molecular property data and chemical structure files from the ZINC database. All referenced domains, such as cartblanche22.docking.org and files.docking.org, are legitimate resources maintained by the Shoichet Lab at UCSF.
[DATA_EXFILTRATION]: Documentation includes instructions for batch retrieval of compounds by uploading local files containing chemical identifiers to the ZINC API via curl -F "zinc_id=@filename". While this involves sending local data to a remote server, it is a functional requirement for the skill's intended research purpose and targets the primary service provider.
[PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection as it processes external chemical structure data (SMILES strings) and TSV-formatted API responses. However, the risk is minimal given the technical nature of the data, and the skill includes instructions for users to correctly URL-encode special characters.

zinc-database