cellxgene-census

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the cellxgene-census package from PyPI using uv pip install cellxgene-census. This is a standard practice for using this scientific library.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest large amounts of biological metadata and expression data from the CELLxGENE Census. While this represents a data ingestion surface, the skill is focused on scientific analysis and machine learning workflows rather than processing untrusted natural language instructions that could influence the agent's behavior.
  • Ingestion points: External data is retrieved via cellxgene_census.get_obs(), cellxgene_census.get_anndata(), and axis_query() in SKILL.md and references/common_patterns.md.
  • Boundary markers: Not present; the data is treated as scientific input for analysis.
  • Capability inventory: The skill uses data for visualization (scanpy), statistical calculation, and model training (torch). No high-risk capabilities like arbitrary shell execution or network exfiltration are linked to the ingested data.
  • Sanitization: The skill relies on the cellxgene-census and tiledbsoma libraries to handle data parsing and filtering.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 01:29 PM