alterlab-geniml

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly ingests public, user-generated genomic data via BBClient (references/utilities.md shows client.load_bed(bed_id='GSM123456') fetching BED files from remote BEDbase/GEO-like repositories) and references pulling pre-trained models from Hugging Face, so untrusted third‑party content is read and used to drive tokenization, training, searches, and downstream actions.