scanpy

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

• Third-party content exposure detected (medium risk: 0.30). Outsider-authored free text can be ingested as JSON/CSV mapping parameters to scripts/annotate.py (--mapping and optional --markers), which the workflow directly parses (json.load/pandas.read_csv) and then uses to generate labels/plots without selecting a specific trusted item first.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The agent runbook (references/r_interop.md) includes R install commands that set CRAN to "https://cloud.r-project.org" and uses remotes::install_github / references to GitHub repos (e.g. https://github.com/mojaveazure/seurat-disk and https://github.com/theislab/zellkonverter), which will fetch and install/execute remote package code at runtime when converting R objects to .h5ad.