insight-skill-gap

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

• Third-party content exposure detected (medium risk: 0.65). The required workflow reads “recent context” including “changed files” and “new patterns introduced” (Phase 1, step:gather), which at runtime can include outsider-authored free text from public/other parties’ contributions (e.g., PR/issue/wiki/discussion bodies) that the agent then summarizes and feeds back into the LLM context.