lov-auto-context

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

• Third-party content exposure detected (medium risk: 0.30). In Auto Mode (skill-publisher UserPromptSubmit hook), the agent evaluates the current user/host transcript size and content and then may auto-write project memory or request diffs based on that read text, meaning outsider-authored free text from the monitored prompt/transcript stream is ingested at runtime.