testing-integration

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

• Third-party content exposure detected (high risk: 0.80). The skill's golden dataset (backend/data/golden_dataset_backup.json) lists external URLs and the "Dataset Backup & Restore" / "Why backup? ... Faster than re-analyzing 98 URLs" instructions show restoring regenerates embeddings (re‑analyzing those URLs), which clearly requires fetching and ingesting public third‑party web content that the tests/retreival/evidence workflows will read and use to influence outcomes.