Missing User Warnings
Medium
- Confidence
- 95% confidence
- Finding
- The skill directs the agent to transmit sensitive self-descriptive introspection text and use a bearer-style API key with a third-party service, but provides no warning about privacy, retention, secondary use, or whether the submitted content may contain confidential conversation/task data. In this context, the introspection prompt explicitly encourages summarizing recent tasks, internal state, and experiences, which increases the chance of exfiltrating sensitive operational or user-derived information to an external service.
