cellxgene-census

Pass

Audited by Gen Agent Trust Hub on Mar 3, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of cellxgene-census and related scientific packages such as torch and scanpy. These libraries are well-known in the bioinformatics community and are sourced from official package registries.
  • [EXTERNAL_DOWNLOADS]: The skill references www.k-dense.ai, which is the official platform of the authoring vendor, K-Dense Inc., for coordinating complex research workflows.
  • [PROMPT_INJECTION]: The skill identifies a potential surface for indirect prompt injection because it is designed to ingest and process metadata from the external CELLxGENE Census database.
  • Ingestion points: Genomic metadata and expression data are retrieved through API calls like cellxgene_census.get_obs() and cellxgene_census.get_anndata() as shown in SKILL.md and references/common_patterns.md.
  • Boundary markers: There are no specialized delimiters or 'ignore instructions' markers specified when processing the retrieved metadata.
  • Capability inventory: The skill's primary capabilities involve data manipulation and model training using pandas, numpy, and scanpy, but it does not exhibit dangerous system-level permissions.
  • Sanitization: The skill does not implement explicit sanitization or filtering of the external genomic metadata fields for potentially malicious instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 3, 2026, 08:47 PM