scvi-tools

Pass

Audited by Gen Agent Trust Hub on Aug 17, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the scvi-tools package using the uv package manager. Additionally, functions like scvi.data.add_dna_sequence are documented to download genomic reference data from external repositories when install_genome=True is used.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process various single-cell genomics data formats, creating a surface for indirect prompt injection.
  • Ingestion points: Data is loaded from external files using scanpy functions such as read_h5ad, read_10x_mtx, read_csv, and read_visium in references/workflows.md and references/models-spatial.md.
  • Boundary markers: The instructions lack explicit boundary markers or directives for the agent to ignore potential instructions embedded within the metadata of processed genomic datasets.
  • Capability inventory: The skill includes commands to write data and models to the local file system using model.save() and adata.write() as shown in references/workflows.md and references/models-scrna-seq.md.
  • Sanitization: There is no evidence of specific sanitization or validation logic for the content of the ingested genomics files before they are processed by the generative models.
  • [COMMAND_EXECUTION]: The skill utilizes shell commands for environment setup and package management, specifically using uv pip install to manage Python dependencies.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 17, 2026, 02:13 AM