scvi-tools

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFE
Full Analysis
  • COMMAND_EXECUTION (SAFE): The skill contains standard installation commands (pip install scvi-tools) and data processing scripts. These are consistent with the skill's stated purpose of providing documentation for a scientific library.
  • INDIRECT_PROMPT_INJECTION (LOW): The skill demonstrates functions that load data and models from local paths (scvi.model.SCVI.load). While these are potential ingestion points for untrusted data, they represent standard library usage for the intended domain.
  • Ingestion points: scvi.data.heart_cell_atlas_subsampled(), scvi.model.SCVI.load("./model_directory") (File: SKILL.json)
  • Boundary markers: Absent (standard for code snippets)
  • Capability inventory: pip install, model.train(), model.save() (File: SKILL.json)
  • Sanitization: Not applicable for these scientific API calls
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:09 PM