tooluniverse-spatial-transcriptomics

Pass

Audited by Gen Agent Trust Hub on Mar 8, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill defines a structured workflow for analyzing spatial transcriptomics data. All Python code snippets provided use legitimate, well-known open-source scientific libraries such as Scanpy, Squidpy, and Scikit-learn for data manipulation and visualization.
  • [PROMPT_INJECTION]: The skill includes functions to ingest external datasets in Phase 1 (e.g., 10x Visium, MERFISH). The data ingestion points are restricted to structured scientific data formats processed by standard parsers (e.g., sc.read_visium). No mechanisms for executing instructions embedded within the data or bypassing agent safety filters were identified.
  • [EXTERNAL_DOWNLOADS]: The skill references the OmniPath database for ligand-receptor interactions. This interaction is facilitated via an internal tool call (tu.run_one_function) to a biological reference service, which is a standard and safe operation in bioinformatics contexts.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 8, 2026, 01:05 PM