nanogpt

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

• Third-party content exposure detected (high risk: 0.90). The skill's data preparation scripts shown in references/data.md and SKILL.md (e.g., data/shakespeare_char/prepare.py which uses requests.get to pull raw text from GitHub and data/openwebtext/prepare.py which loads the public OpenWebText dataset via datasets.load_dataset) explicitly fetch and ingest untrusted public/web user-generated content that is then used for tokenization and model training, so that third-party text can materially influence the model's behavior.