weekly-review

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

• Third-party content exposure detected (high risk: 0.90). Yes—at runtime the skill’s LLM context will include hostile free-form content from the Gmail connector (email body text and potentially attachments/transcriptions treated as “email body”), used in “Scan the inbox for open loops” and then surfaced in the delivered review (“Open loops (inferred from inbox…)”).