sentiment-monitor

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

• Third-party content exposure detected (high risk: 0.85). The skill’s workflow calls Promptwatch MCP tools (e.g., list_responses, get_response_sentiment_time_series) that return “response excerpts”/text from responses stored in Promptwatch; those response bodies are outsider-authored free-form content (not written by the operating user) and are ingested into the agent’s LLM context via the MCP tool outputs.