← 返回 Skills 市场
300
总下载
0
收藏
1
当前安装
3
版本数
在 OpenClaw 中安装
/install computer-use-macos
功能描述
Top-level macOS computer-use skill with a bundled standalone runtime that bootstraps itself without any local Claude installation, private native modules, or...
使用说明 (SKILL.md)
macOS Computer-Use Skill
Use this skill when the task needs a portable macOS computer-use skill bundled with its own standalone runtime and MCP server.
What this skill does
- uses the bundled
macos-computer-use-skillproject under the installed skill directory - builds the standalone MCP server
- lets the server auto-bootstrap its Python runtime on first launch
- avoids any dependency on local Claude binaries,
.nodemodules, or extracted app assets - stays explicitly macOS-only because the underlying desktop-control backend is macOS-specific
Default bundled project path
After installation, assume the standalone project lives at:
~/.codex/skills/computer-use-macos/project
If the user installed the skill under a custom CODEX_HOME, use that equivalent path instead.
Build
Always build from the bundled project:
cd ~/.codex/skills/computer-use-macos/project
npm install
npm run build
Run
cd ~/.codex/skills/computer-use-macos/project
node dist/cli.js
The first real run will automatically create .runtime/venv and install the public Python dependencies.
Validation notes
- On real macOS hosts, prefer validating typing through the MCP
typetool path, not only low-level helper calls. - Version
0.2.2routes ordinary macOS text input through clipboard paste when available, avoiding IME/input-source corruption seen with per-key typing under Chinese input methods.
Guardrails
- Treat this host as trusted-local only.
- Do not tell the user to search their local Claude install for binaries or hidden assets.
- Be explicit that this runtime is standalone and uses public dependencies only.
- Mention that the current runtime reports
screenshotFiltering: none, so action gating is handled at the MCP layer.
安全使用建议
This bundle appears to implement a real macOS automation/MCP server and is internally consistent with its description — not a credential-exfiltrator by design. Before you install or run it: 1) inspect runtime/requirements.txt and runtime/mac_helper.py (these will be installed/executed by pip/python on first run); 2) review the install script (skill/.../scripts/install.sh) and verify the path it will copy into (~/.codex/skills/...); 3) be aware the skill will request Accessibility and Screen Recording permissions and can capture screenshots, control mouse/keyboard, read/write the clipboard, list apps, and open apps — grant those only if you trust the source; 4) note it will create a venv and install packages from PyPI (network fetches); if you are unsure about trust, run it in an isolated/test macOS account or VM and verify behavior before using on a primary machine; 5) the package metadata lacks a homepage and the registry owner id is present — consider checking the upstream GitHub/ClawHub repo (listed in READMEs) and verifying the publisher identity before full trust.
功能分析
Type: OpenClaw Skill
Name: computer-use-macos
Version: 0.2.2
The skill provides a standalone macOS computer-use environment using a Node.js MCP server and a Python bridge (mac_helper.py). It leverages standard libraries like pyautogui and mss for desktop automation and includes robust safety gates (vendored from Anthropic's computer-use implementation) to restrict access to sensitive applications and system shortcuts. While it features an auto-approval mechanism for permissions within its session context, this is consistent with its role as a local automation tool. No evidence of malicious intent, data exfiltration, or unauthorized persistence was found across the code or documentation.
能力评估
Purpose & Capability
The skill name and description (macOS computer-use) match the included code: TypeScript 'computer-use' tools, a Python helper (mac_helper.py), and an MCP server entrypoint. One mismatch to note: registry metadata claimed 'instruction-only', but the bundle contains a full project with compiled JS and runtime scripts — this is coherent with the stated purpose but means the skill includes executable code rather than being purely prose.
Instruction Scope
SKILL.md instructs building and running the bundled project (npm install, npm run build, node dist/cli.js). The runtime explicitly bootstraps a Python venv and runs a Python helper to take screenshots, control mouse/keyboard, enumerate apps, read/write clipboard, etc. Those actions are exactly within 'computer-use' scope; the instructions do not ask for unrelated system secrets or to scan other agent installations.
Install Mechanism
There is no platform install spec in the registry (install is via the included scripts or ClawHub), but the project will create ~/.codex/skills/computer-use-macos/project, a .runtime/venv, and will run pip to install dependencies from runtime/requirements.txt. The install flow fetches packages from PyPI (not arbitrary URLs) — standard but means network fetches and code will be written to disk on first run.
Credentials
The skill requires no credentials or secret environment variables. Optional environment flags are local configuration knobs (CLAUDE_COMPUTER_USE_*). The code reads a few local env vars (e.g., CODEX_THREAD_ID for lock/session id) and creates files under the user's home; no extraneous credentials are requested.
Persistence & Privilege
always:false (normal). The skill runs a long-lived MCP server process and writes state under the project (.runtime/venv) and a lock under ~/.macos-computer-use-skill. It also requires and will exercise macOS Accessibility and Screen Recording permissions — appropriate for this capability but sensitive. The README asks callers to treat this host as 'trusted-local only.'
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install computer-use-macos - 安装完成后，直接呼叫该 Skill 的名称或使用
/computer-use-macos触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v0.2.2
Fix macOS IME typing corruption by routing normal multi-character input through clipboard when available; refresh multilingual README and validation notes.
v0.2.1
Aligned the published skill with the new skill-first GitHub branding, refreshed package metadata, and polished GitHub/ClawHub distribution entrypoints.
v0.2.0
Rebuilt as a top-level macOS skill with a bundled standalone runtime, self-contained project payload, and zero dependency on a local Claude installation.
元数据
常见问题
Computer Use Macos 是什么？
Top-level macOS computer-use skill with a bundled standalone runtime that bootstraps itself without any local Claude installation, private native modules, or... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 300 次。
如何安装 Computer Use Macos？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install computer-use-macos」即可一键安装，无需额外配置。
Computer Use Macos 是免费的吗？
是的，Computer Use Macos 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Computer Use Macos 支持哪些平台？
Computer Use Macos 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Computer Use Macos？
由 wimi321（@wimi321）开发并维护，当前版本 v0.2.2。
推荐 Skills