email-drafting

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly ingests and interprets third-party, user-generated content — it reads incoming Gmail messages to detect triggers and uses Grain transcripts via mcporter (see "Trigger Detection" and "Use Grain as primary source for meeting-based drafts") to draft emails, so external participants' text can directly influence agent decisions and tool actions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill calls mcporter to fetch Grain meeting transcripts at runtime (see https://grain.com/docs), and those external transcripts are injected into the drafting flow and directly control the agent's generated email content.