engage

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

• Third-party content exposure detected (high risk: 1.00). The skill's Step 1 explicitly calls list_engagement to pull unread comments/mentions/replies from social platforms (user-generated, public third-party content) and then reads and acts on those items to triage and generate/send replies (SKILL.md), so untrusted external content can influence agent decisions and actions.