baseline-models

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

• Third-party content exposure detected (medium risk: 0.30). In scripts/run_baselines.py and scripts/home_only_baseline.py, the agent reads user-supplied tabular files via the --input path at runtime and processes the contained columns/values as feature/target data (so outsider-authored free text can be ingested if it is placed into those files), meaning the workflow is exposed to poison through the input dataset.