云手机运管平台操作

火山引擎云手机运管平台操作技能。支持快速打开运管平台、搜索实例、打开实例详情、下载文件等操作。使用场景：(1) 用户说"打开云手机运管平台"；(2) "搜索实例 123456"；(3) "下载实例日志"。

This skill appears to do what it says (drive a browser to your internal vcloud admin UI and help you download instance logs), but take these precautions before installing or using it: - Confirm you are authorized to access the internal domain (https://vcloud-admin.bytedance.net). The skill assumes an authenticated browser session; it will not provide credentials. - Be careful which files you download and copy into the workspace: instance logs can contain sensitive information. Only copy files you intend the agent to read. - Verify the workspace path (/Users/bytedance/.openclaw/workspace/) matches your environment; if not, change it or refuse to copy sensitive files there. - Because browser download dialogs and OS-level UI require manual interaction, expect the skill to prompt you for steps rather than fully automate them. - If you have concerns about exposing production data, perform these steps on a controlled/test instance or scrub logs before copying them into the workspace. Confidence is medium because the skill targets an internal service and has no source/homepage; these are explainable but worth verifying with your organization before use.

Type: OpenClaw Skill Name: vcloud-phone Version: 1.0.0 The skill bundle is suspicious because it instructs the agent to facilitate the download of arbitrary file paths from a remote 'cloud phone' platform (e.g., '/data/misc/logd/logcat' or user-specified paths) via browser automation. While the skill itself does not directly exfiltrate data, it instructs the user to manually copy these potentially sensitive downloaded files into the agent's local workspace (`/Users/bytedance/.openclaw/workspace/`). The skill then explicitly instructs the agent to 'read' these files from the workspace, leveraging its declared 'Read' tool capability. This combination exposes a high-risk capability for data leakage from the remote system, making sensitive data accessible to the agent for potential further processing, even without explicit malicious intent in the current instructions.