source-credibility-evaluation-protocol

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

• Third-party content exposure detected (medium risk: 0.65). The skill’s runtime prompt generation can include a “Specific source” and then instructs the teacher to model lateral reading by searching external information about that source (e.g., “search the author/organisation name” and “find better coverage”), which at runtime can pull in outsider-authored free text from public web pages into the agent’s LLM context via the browsing/search results.