skills/d-laub/genoray/genoray-api/Gen Agent Trust Hub

genoray-api

Pass

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill enables processing of external genomic files (VCF, PGEN, SparseVar), which creates an attack surface for indirect prompt injection if the processed data contains malicious instructions.
  • Ingestion points: External bioinformatics data files as described in SKILL.md.
  • Boundary markers: The API processes structured binary and text formats but lacks specific delimiters or instructions for the agent to ignore embedded Natural Language instructions within those files.
  • Capability inventory: File system read/write access, network data fetching for reference signatures, and parallel task execution.
  • Sanitization: Relies on established parsing libraries such as cyvcf2 and pgenlib.
  • [EXTERNAL_DOWNLOADS]: The genoray.cosmic_signatures method fetches scientific reference data from the COSMIC database. This is a standard operation targeting a well-known research service and is consistent with the skill's bioinformatics purpose.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 21, 2026, 04:09 PM