alphagenome
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external genomic data files (VCF, TSV, CSV) for variant analysis, creating a standard attack surface for indirect prompt injection via data ingestion.
- Ingestion points: Genomic data is ingested from local files through the
--inputargument in the analysis scripts (scripts/atlas_query.pyandscripts/score_variants.py). - Boundary markers: The scripts utilize structured parsing for VCF and delimited formats, validating coordinate and base formatting.
- Capability inventory: The skill has access to the
Bashtool for environment setup, can write results to the filesystem, and performs network requests to AlphaGenome gRPC API endpoints. - Sanitization: Genomic variants are validated against expected regex patterns and reference alleles, ensuring data integrity before processing.
- [EXTERNAL_DOWNLOADS]: The skill fetches reference genome annotations from Google Cloud Storage (
storage.googleapis.com) and installs the corealphagenomepackage from PyPI. These resources originate from a well-known and trusted service provider.
Audit Metadata