geniml

Pass

Audited by Gen Agent Trust Hub on Nov 6, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external genomic data and metadata, which could be manipulated to include hidden instructions. * Ingestion points: The skill ingests BED files, metadata CSVs, and remote datasets via the BBClient caching utility. * Boundary markers: There are no explicit instructions to the agent to treat genomic annotations or metadata values as untrusted or to ignore embedded commands. * Capability inventory: The skill has the ability to read/write files and execute command-line tools based on the processed data. * Sanitization: The instructions do not specify validation or sanitization steps for the contents of the genomic files or metadata fields.
  • [EXTERNAL_DOWNLOADS]: The skill fetches code and models from various external repositories. * It installs the geniml package from official Python registries. * It provides instructions to install development versions from the databio GitHub organization. * It references the StarSpace tool from the well-known facebookresearch GitHub organization. * It downloads pre-trained models from the Hugging Face model registry.
  • [COMMAND_EXECUTION]: The skill uses several command-line utilities to perform its tasks. * It executes the geniml CLI for model training, preprocessing, and searching. * It uses bioinformatics tools like uniwig and standard shell commands like cat for data manipulation.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 6, 2026, 12:15 PM