gtars

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFE
Full Analysis
  • Indirect Prompt Injection (SAFE): The skill processes external genomic files (BED, FASTA, TSV) which are untrusted data sources. Ingestion points: RegionSet.from_bed, RefgetStore.from_fasta, and TreeTokenizer.from_bed_file. Boundary markers: None explicitly defined in documentation. Capability inventory: File processing, interval arithmetic, and tokenization. Sanitization: Standard bioinformatics format parsing. While this represents a data ingestion surface, the processing is mathematical/structural rather than instructional, making the risk negligible.
  • Unverifiable Dependencies & Remote Code Execution (SAFE): The skill recommends installing 'gtars' and 'geniml' via standard package managers like pip, uv, and cargo. These follow common distribution patterns for bioinformatics libraries. No suspicious download-to-execution pipelines (like curl | bash) were identified.
  • External Downloads (SAFE): The 'bbcache' module includes functionality to fetch genomic data from bedbase.org. This is a domain-specific feature for caching reference datasets from a known public repository.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 05:02 PM