Natural-Language Policy Violations
Medium
- Confidence
- 89% confidence
- Finding
- The skill explicitly marks Chinese triggers as 'preferred' without any user opt-in, which can cause the agent to bias language selection and routing behavior based on embedded skill instructions rather than the user's stated preference. While not a code-execution issue, this can override expected interaction boundaries and create unfair or confusing behavior for users who did not request Chinese-language prioritization.
