geniml
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- [Unverifiable Dependencies & Remote Code Execution] (SAFE): The skill references the external dependency StarSpace and pre-trained models from Hugging Face. Both are hosted by trusted entities (Facebook Research and Hugging Face, respectively). There are no commands for automated, piped, or unauthorized remote script execution.\n- [Data Exposure & Exfiltration] (SAFE): All documented file operations pertain to the processing of genomic datasets. The BBClient utility uses a standard local caching mechanism (~/.bedcache) which does not target sensitive user configurations or credentials.\n- [Indirect Prompt Injection] (SAFE): The skill ingests external genomic and metadata files for training and search tasks. These operations are performed within the geniml framework and do not expose sensitive execution sinks to the data content, effectively mitigating the risk of indirect instruction injection.
Audit Metadata