The Agent Skills Directory

[SAFE]: No prompt injection or behavior override instructions were found in the skill metadata or body. The skill maintains clear boundaries and follows standard research protocols.
[DATA_EXFILTRATION]: Data access is limited to well-known scientific and medical databases (e.g., DrugBank, ChEMBL, PubMed, FDA). No sensitive system files are accessed, and no exfiltration to suspicious domains occurs.
[EXTERNAL_DOWNLOADS]: All external references and tool calls target established scientific organizations and government databases. Provided URLs point to official academic domains (zitniklab.hms.harvard.edu) associated with the authoring institution (Harvard Medical School).
[COMMAND_EXECUTION]: Python examples demonstrate legitimate use of a research SDK without involving dangerous system commands or arbitrary code execution using functions like eval() or exec().
[SAFE]: No obfuscation techniques, hidden characters, persistence mechanisms, or privilege escalation attempts were detected.

tooluniverse-drug-repurposing