tooluniverse-spatial-transcriptomics

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

• Third-party content exposure detected (high risk: 0.70). The skill explicitly calls external ToolUniverse APIs to fetch public third‑party data (e.g., Phase 7.5 "OmicsDI_search_datasets/NCBI_SRA_search_runs/UniProt/STRING/KEGG/DGIdb/PubMed" and the code example's tu.run_one_function("OmniPath_get_ligand_receptor_interactions")), and those returned open/public dataset, literature and ligand–receptor lists are ingested and used to drive analysis and scoring (e.g., ligrec), so untrusted third‑party content can materially influence tool behavior.