bio-protein-clustering-pangenome
Pass
Audited by Gen Agent Trust Hub on Feb 19, 2026
Risk Level: SAFE
Full Analysis
- [Prompt Injection] (SAFE): No instructions were found that attempt to override the agent's system prompt, bypass safety guidelines, or extract sensitive internal configurations.
- [Data Exposure & Exfiltration] (SAFE): There are no hardcoded credentials, API keys, or suspicious network operations. File access is limited to the expected input and output directories for biological sequence analysis.
- [Remote Code Execution] (SAFE): Installation steps for external tools (MMseqs2, OrthoFinder, etc.) use trusted package managers and official source repositories. No unverified or piped execution of remote scripts was identified.
- [Indirect Prompt Injection] (LOW): The skill processes protein FASTA files which are untrusted external inputs. However, these are processed by compiled binaries (MMseqs2) and the workflow includes a quality gate to verify the file encoding, minimizing the risk of the LLM interpreting sequence data as instructions.
Audit Metadata