llm-finetuning

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

• Third-party content exposure detected (high risk: 0.90). The skill's runnable notebooks explicitly download and ingest public, user-generated content — e.g., datasets.load_dataset("ag_news") in scripts/demo.py and the FLAT_CMPL.zip from https://static.nhtsa.gov/... in scripts/demo_nhtsa.py — and the workflow requires the agent to read and act on that content (formatting, training, and inference), so untrusted third-party text can materially influence model behavior and subsequent actions.