bio-expression-matrix-sparse-handling

Warn

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: MEDIUMCOMMAND_EXECUTION
Full Analysis
  • [Dynamic Execution] (MEDIUM): Detected unsafe deserialization pattern in Python code.
  • Evidence: loaded = np.load('counts_with_meta.npz', allow_pickle=True) in file SKILL.md.
  • Risk: The pickle module is not secure against erroneous or maliciously constructed data. Enabling allow_pickle when loading data from potentially untrusted sources can lead to arbitrary code execution on the host machine.
  • [Indirect Prompt Injection] (LOW): The skill possesses an ingestion surface for untrusted external data.
  • Ingestion points: File reading operations in SKILL.md including pd.read_csv('counts.csv'), scipy.io.mmread('matrix.mtx'), and ad.read_h5ad('counts.h5ad').
  • Boundary markers: None. Data is loaded directly into memory objects.
  • Capability inventory: Matrix operations, file read/write. No network operations or shell access capabilities were detected in the scripts.
  • Sanitization: No validation or sanitization of file contents is performed before processing.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 16, 2026, 05:46 AM