Academic Dataset Selection and Analysis

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md explicitly requires the agent to "find a dataset from the internet" and in Operational Rules/Interaction Workflow (e.g., "Must be from the internet with a citable source (URL)" and "Find and present a dataset... including source URL") the agent is required to fetch and interpret arbitrary public datasets, exposing it to untrusted third-party content that could carry injected instructions.