brand-voice

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

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime workflow explicitly ingests “recent original X posts and threads” via x-api (outsider-authored public social content) and/or “articles… newsletters… real outbound emails or DMs that worked” (messages authored by non-operating-user parties), which can be read as free-form text and fed into the agent to extract a voice profile.