查看B站UP主的最新视频、动态，检查UP主今天是否更新。触发词：B站、UP主、视频更新、今天更新了吗、最新视频、最新动态、查看UP主

This skill appears to do what it says (query Bilibili APIs for UP updates) and only communicates with Bilibili endpoints, but it requires your BILIBILI_COOKIES (session cookie) to work. Before installing or running it: - Be cautious with session cookies: providing SESSDATA or full cookies gives code the ability to act as your account. Only supply cookies if you trust the code and maintainer. Consider using a disposable account or a cookie that has minimal privileges. - The registry metadata did NOT declare the required BILIBILI_COOKIES env var — ask the publisher to add it to the manifest so permission requests are transparent. - Inspect the included Python files locally (they are present in the package). Confirm the scripts only call api.bilibili.com / bilibili.com (they do) and do not post data to other hosts. - Expect a user_cache.json file to be created next to the scripts; it contains cached username→mid entries. If you do not want on-disk cache, run the scripts in a disposable directory or remove cache-writing calls. - Run the tool in an isolated environment (virtualenv or container) and avoid uploading your real account cookie to remote or shared environments. If the publisher updates the registry metadata to declare BILIBILI_COOKIES explicitly and you validate the code locally, the remaining risks are normal for a tool that needs a site cookie.

Type: OpenClaw Skill Name: bilibili-update-viewer Version: 1.0.0 The skill bundle is classified as suspicious due to a critical shell injection vulnerability. The `SKILL.md` file instructs the OpenClaw agent to execute Python scripts by directly embedding user-provided input (e.g., `用户名`) into shell commands without explicit sanitization (e.g., `python3 {baseDir}/get_mid.py "用户名"`). If the agent does not properly escape or sanitize this input before execution, an attacker could inject arbitrary shell commands, leading to Remote Code Execution (RCE). While the Python code itself appears benign and focused on its stated purpose, this RCE risk stemming from the agent's execution instructions makes the skill suspicious.