camp-profile

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

• Third-party content exposure detected (high risk: 0.75). The required workflow explicitly has the agent fetch and read the camp’s own website/brochure at runtime (“Start from the website” → “ask for the camp's website URL… If you can read it, do: pull…”), which is public web content authored by an outsider, and that readable text can flow into the LLM context via the fetched page content.