anndata

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.70). The skill's IO documentation (SKILL.md and references/io_operations.md) explicitly shows reading from arbitrary URLs and remote stores (e.g., ad.read_h5ad(url, backed='r'), fsspec.get_mapper('https://...') / S3 Zarr access and the "Download from URL" example), so the agent can fetch and interpret untrusted public third‑party content that could materially influence subsequent processing steps.