geniml

Pass

Audited by Gen Agent Trust Hub on May 3, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the user to install the 'geniml' package via standard package managers and provides links to reputable external resources, including official GitHub repositories from the 'databio' research group and 'facebookresearch' (for StarSpace). It also references pre-trained models hosted on Hugging Face.
  • [COMMAND_EXECUTION]: The skill uses standard CLI tools and subcommands (geniml, uniwig, uv) to process genomic data. These operations are restricted to the intended purpose of the skill and do not involve shell injection or execution of untrusted scripts.
  • [DATA_EXFILTRATION]: No suspicious network operations were identified. The BBClient utility is used for local caching of genomic data (~/.bedcache) to improve performance, which is a standard practice in bioinformatics workflows.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data such as BED files and CSV metadata. While this represents a potential ingestion surface for indirect instructions, the skill uses specialized parsers (like pandas and geniml's tokenizers) and limits its capabilities to data processing and model training, posing no significant security risk in its current form.
Audit Metadata
Risk Level
SAFE
Analyzed
May 3, 2026, 11:19 PM