Security audit
Security checks for vulnerabilities and agentic risk
This skill is coherent for an internal cloud-phone admin workflow, but it gives broad production data access through browser actions and arbitrary instance file downloads without enough scoping or confirmation.
Install only if users are authorized production operators for this admin platform. Before use, require the agent to confirm the exact instance ID and file path, avoid arbitrary filesystem downloads unless explicitly approved, and treat copied downloads as potentially sensitive production data.
The skill description, triggers, warnings, and examples are entirely in Chinese, with no indication that users may interact in another language or that Chinese is required for a region-specific reason. This creates a language/locale policy concern because the skill implicitly constrains interaction language without opt-in.
The trigger phrase for opening instance details is context-poor and can fire on a vague follow-up like '打开实例详情' without verifying that a prior search selected the intended instance. In an admin console for a production cloud-phone platform, this can cause the agent to navigate to the wrong tenant/resource and expose sensitive operational data or set up later unsafe actions.
The download triggers are broad enough to authorize arbitrary file-path downloads from an instance, including direct path-based requests such as '/data/misc/logd/logcat' or other user-supplied paths, without clear scope restrictions or approval gates. In the context of a production operations platform, this creates a real risk of unauthorized retrieval of sensitive logs, device data, secrets, or other filesystem contents from the wrong instance.
No suspicious patterns detected.
