persona-review

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

• Third-party content exposure detected (high risk: 0.80). The required workflow reads a user-supplied document from file-path (Step 1) and then feeds its extracted frontmatter/filename/content as text into the agent LLM context for persona reviews, making the document body an outsider-authored free-text source if the operating user didn’t author it.