The Agent Skills Directory

[SAFE]: No malicious patterns or security vulnerabilities were detected in the skill files.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface for molecular formats (SMILES, PDB, XYZ) which could be used for indirect prompt injection. However, given the skill's specific purpose and lack of high-risk capabilities, the risk is minimal.
Ingestion points: Data is loaded via stjames.Molecule.from_smiles, stjames.Molecule.from_file, and rowan.upload_protein across SKILL.md and reference files.
Boundary markers: None identified.
Capability inventory: Limited to specialized chemistry API calls and downloading result files (SDF, PDB) to the local system.
Sanitization: Includes chemical structure validation through the stjames library but lacks specific adversarial instruction filtering.
[DATA_EXPOSURE]: The skill manages authentication through API keys and correctly recommends using environment variables for secure storage, consistent with best practices.
[EXTERNAL_DOWNLOADS]: All network operations are directed to the Rowan platform's official infrastructure at labs.rowansci.com and the author's site k-dense.ai.

rowan