investigate-errors

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

• Third-party content exposure detected (high risk: 0.70). The skill fetches “recent error issues” and “stack traces and context” from Sentry via MCP tools at runtime; those issue/event texts are authored/recorded by third parties (i.e., Sentry-ingested application error data) rather than by the operating user, so the LLM ingests outsider-origin free text through the Sentry MCP retrieval step.