check-inbox

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

• Third-party content exposure detected (high risk: 0.90). This skill explicitly ingests untrusted, user-generated messages from local Gmail markdown files (./output/gmail/) and Slack exports (./slack-sync/data/) and passes their sanitized contents into scripts/llm_analyzer.py for LLM analysis and reply-draft generation, so third-party message text can materially influence actions despite boundary/tagging mitigations.