The Agent Skills Directory

[SAFE]: The skill implements medicinal chemistry workflows using the medchem and datamol libraries. All operations, such as calculating Lipinski rules or filtering structural alerts (PAINS), are standard scientific data processing tasks with no security implications.
[SAFE]: The installation instructions use standard package managers (uv pip install medchem) targeting a legitimate open-source library maintained by a recognized organization in the chemoinformatics field.
[SAFE]: The provided Python script scripts/filter_molecules.py is a well-structured command-line utility for batch processing molecular files (CSV, SDF, TXT). It performs local file I/O and parallel computation without any network exfiltration, dynamic code execution, or privilege escalation patterns.
[SAFE]: Metadata and documentation are consistent with the skill's stated purpose, with no signs of prompt injection or metadata poisoning.
[SAFE]: The skill identifies a potential attack surface for Indirect Prompt Injection (Category 8) as it processes external chemical data files. However, it implements robust boundary handling by converting raw SMILES strings into structured RDKit molecule objects and discarding invalid inputs before analysis, effectively mitigating risk.

medchem