alterlab-gtars

Pass

Audited by Gen Agent Trust Hub on Apr 12, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill facilitates downloading genomic data from Bedbase (bedbase.org), a well-known bioinformatics repository, through the gtars bbcache fetch command.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface due to its processing of external genomic files.
  • Ingestion points: Genomic files are ingested via RegionSet.from_bed, RefgetStore.from_fasta, and TreeTokenizer.from_bed_file across various reference modules.
  • Boundary markers: The skill does not employ specific boundary markers or instructions to isolate processed data from the agent's control flow.
  • Capability inventory: The skill performs file system operations (read/write) and executes shell commands for genomic analysis via the gtars CLI.
  • Sanitization: No explicit sanitization or content validation is performed on the ingested genomic data.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 12, 2026, 12:46 AM
Security Audit — agent-trust-hub — alterlab-gtars