dataset-transformation

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

• Third-party content exposure detected (high risk: 0.80). The skill's SKILL.md Evaluation Data Formats step explicitly instructs the agent to fetch live documentation from the public URL https://docs.aws.amazon.com/sagemaker/... to resolve evaluation dataset schemas and to "use whichever source you successfully access as the source of truth," meaning external web content is read and used to drive transformation decisions.