model-training

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

• Third-party content exposure detected (high risk: 0.80). The skill's workflow explicitly loads datasets from remote/public sources (e.g., "Load and inspect data" and Example 2's use of datasets.load_dataset("imdb") from the Hugging Face Hub), meaning the agent ingests untrusted, user-generated third-party content that can directly influence training decisions and downstream behavior.