anndata

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADS
Full Analysis
  • [EXTERNAL_DOWNLOADS] (LOW): The documentation includes examples for downloading and reading data from external network sources.
  • Evidence in references/io_operations.md: Uses urllib.request.urlretrieve(url, 'local_file.h5ad') and fsspec.get_mapper('https://example.com/data.zarr') to access remote data stores.
  • [Indirect Prompt Injection] (LOW): The skill identifies a significant data ingestion surface that could be exploited if an agent processes untrusted scientific data files.
  • Ingestion points: ad.read_h5ad(), ad.read_csv(), ad.read_excel(), ad.read_zarr(), and ad.read_mtx() in references/io_operations.md.
  • Boundary markers: None documented; the library reads data structures directly into memory.
  • Capability inventory: Subprocess-like capability is absent, but the skill enables extensive file-system writes (write_h5ad, write_csvs, write_zarr) and network reads.
  • Sanitization: Not present; the library assumes data integrity of the input files.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 05:00 PM