Missing User Warnings
Medium
- Confidence
- 98% confidence
- Finding
- The skill explicitly routes a user's task to multiple third-party LLMs and then forwards their outputs to a merge model, but the description does not warn users that their prompts and intermediate model responses may be disclosed to several external providers. This creates a real confidentiality and compliance risk because sensitive data may be replicated across multiple vendors, jurisdictions, retention policies, and training pipelines without informed consent.
