tooluniverse-multiomic-disease-characterization
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is a purely instructional guide for complex bioinformatics analysis. It uses legitimate external data sources like OpenTargets, GWAS Catalog, and STRING for scientific research purposes.
- [PROMPT_INJECTION]: No patterns of instruction override, jailbreaking, or attempts to bypass safety filters were detected in the instructions.
- [DATA_EXPOSURE]: There is no evidence of hardcoded credentials, sensitive file path access (e.g., .env, .ssh), or unauthorized network exfiltration.
- [REMOTE_CODE_EXECUTION]: The skill does not attempt to install external packages or execute remote scripts via curl/wget piped to a shell.
- [OBFUSCATION]: No obfuscation techniques, including Base64 encoding, zero-width characters, or homoglyph attacks, were identified in the content.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses a typical attack surface as it ingests data from external biological databases (like PubMed and WikiPathways). However, this is required for its primary function, and there are no signs of exploitable patterns combined with this ingestion.
- [DYNAMIC_CONTEXT_INJECTION]: No use of pre-execution shell command patterns was found in the SKILL.md file.
Audit Metadata