tooluniverse-single-cell

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The script scripts/scrna_qc.py employs the import() function to dynamically verify the availability of essential scientific libraries like scanpy, anndata, and numpy. This is used as a pre-flight dependency check to ensure the execution environment is correctly configured before starting data analysis.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied genomics data files, including .h5ad, .csv, and .tsv formats, which represents a potential vector for indirect prompt injection. A malicious actor could attempt to embed instructions within the data content that might influence agent behavior.
  • Ingestion points: Data loading occurs through standard functions like scanpy.read_h5ad and pandas.read_csv in SKILL.md and associated scripts (e.g., scripts/scrna_qc.py, scripts/normalize_data.py).
  • Boundary markers: The skill lacks specific delimiters or instructions designed to isolate data content from the agent's command processing.
  • Capability inventory: The skill possesses the capability to perform file-system writes (e.g., adata.write_h5ad in scripts/qc_metrics.py) and network requests to external biological databases through its integrated toolset.
  • Sanitization: Data is parsed using standard scientific libraries, but there is no additional layer of sanitization for natural language instructions potentially hidden within the data values or metadata.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 04:38 PM
Security Audit — agent-trust-hub — tooluniverse-single-cell