Clash Controller

控制 Clash for Windows 代理：启动、关闭、查看状态、切换节点。触发词：Clash、代理、代理开关、开启代理、关闭代理、代理状态、切换节点、Clash for Windows。

This skill mostly does what it says (controls Clash via the local REST API) but you should be cautious before installing: - The repository contains a hard-coded API secret. That secret is sent to the local controller; remove or replace it and instead configure the skill to read your actual Clash external-controller secret from the config file or from an environment variable. - SKILL.md mentions starting/stopping the Clash process and reading the config file, but the code does not do process control or file reads — expect only REST API calls. If you need process control, request or review additional code. - Test on a non-production machine first: enable External Controller in Clash, verify the controller host/port/secret, and try the skill locally. - If you do not want any reusable tokens stored in the skill, edit skill.js so it reads the secret from an environment variable (or from the local config) rather than using the baked-in token. - Because the skill communicates with a local service using an auth token, ensure the token in your Clash config is private and the machine is secure; the skill does not appear to exfiltrate data externally, but exposing any token in a public registry is a privacy risk. If you want higher assurance, request the author to remove hard-coded secrets and to document exactly how the skill obtains the correct secret (env var or config path) and whether it requires OS-level process control and how that will be implemented.

Type: OpenClaw Skill Name: openclaw-clash-forwindows-controller Version: 1.0.0 The skill bundle is designed to control a local Clash for Windows proxy client via its REST API. The code in `skill.js` performs standard HTTP requests to `127.0.0.1` to toggle proxy settings and list nodes based on user commands. While it contains a hardcoded API secret and references a specific Windows path (`C:\Users\Administrator`), these are functional requirements for interacting with the Clash API and appear to be configuration defaults rather than indicators of malice. No evidence of data exfiltration, remote code execution, or unauthorized system access was found.