The Agent Skills Directory

[SAFE]: No security issues were detected. The skill implements a robust scientific workflow with detailed instructions, quantitative scoring, and evidence grading based on established bioinformatics best practices.
[EXTERNAL_DOWNLOADS]: The skill fetches data from numerous reputable and well-known scientific services, such as OpenTargets, ChEMBL, UniProt, Ensembl, PubMed, and AlphaFold. These references are fundamental to the skill's intended purpose and target official public repositories of scientific data.
[COMMAND_EXECUTION]: Operations are conducted via a controlled set of tool calls within the 'ToolUniverse' platform (using the tu.tools namespace). This approach ensures that the agent's actions are restricted to retrieving and processing scientific data from authorized sources.
[DATA_EXFILTRATION]: There is no evidence of unauthorized access to sensitive system files, environment variables, or credentials. The skill does not attempt to send data to unknown or suspicious external domains.
[PROMPT_INJECTION]: The skill contains no instructions designed to bypass safety filters or override agent behavior. The process of 'Target Disambiguation' in Phase 0 serves as a validation step that mitigates potential risks from malicious user-supplied input.

tooluniverse-drug-target-validation