← 返回 Skills 市场
macOS Desktop Automation (
93
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install application-use
功能描述
Automate macOS tasks by opening apps, clicking elements, filling forms, typing, scrolling, and controlling the desktop via CLI commands.
使用说明 (SKILL.md)
macOS Desktop Automation (application-use)
A CLI to automate macOS applications.
install
npm i -g application-use
🚀 Core Workflow
- Open:
application-use open --appName "\x3CName>"-- Auto-shows initial snapshot. - Interact: Use hints (e.g.,
[A]) to click, type, or scroll. - Auto-Update: Commands (
open,click,type,sendkey,scroll) automatically display an updated snapshot.
[!TIP] Use
application-use snapshotonly when you need a fresh view without any action (e.g., after waiting for dynamic content).
🛠 Command Reference
| Command | Usage | Description |
|---|---|---|
open |
open --appName "Safari" |
Opens app and shows snapshot. |
click |
click A [--right] [--double] |
Clicks element by hint letter. |
fill |
fill A "text" or fill "text" |
Fills text into hint or focused element. |
sendkey |
sendkey enter, sendkey cmd+v |
Sends single or shortcut keys. |
scroll |
scroll "Main" down [500] |
Scrolls area by name/hint. |
screenshot |
screenshot [path] [--frame] |
Captures window or specific coordinates. |
search |
search "Safari" |
Finds installed application names. |
close |
close --appName "Safari" |
Quits the specified application. |
type |
type "text" |
Types text into the focused element. |
🏷 Hint System
- Elements:
[A],[B],[AA]... (Interactive elements). - Areas:
[a],[b]... (Targets for scrolling). - Special Marks:
(*)= OCR-detected;(+)= Newly added since last snapshot.
⚡️ Common Automation Patterns
Form Filling & Submission & scroll
application-use open --appName "Safari"
application-use fill A "[email protected]" # Fill email
application-use sendkey tab # Next field
application-use fill "password" # Focus-fill password
application-use sendkey enter # Submit
application-use scroll d down 1000 # Scroll down area d
Browser Navigation
application-use open --appName "Safari"
application-use fill A "https://google.com" && application-use sendkey enter
# Wait for load, then interact with hints from auto-snapshot
File Operations (Finder)
application-use open --appName "Finder"
application-use sendkey cmd+shift+g # Go to Folder
application-use fill "/Users/name/Downloads" && application-use sendkey enter
# Open in New Tab via right-click
application-use click A --right # Right-click folder/file (hint A)
application-use click B # Click "Open in New Tab" from menu (hint B)
安全使用建议
This skill's purpose (macOS automation) is plausible, but there are red flags you should address before installing:
- Verify provenance: the registry provides no homepage/source. Search npm for 'application-use', inspect the package author, recent versions, and read the package source before installing. Prefer installing only from trusted, audited packages.
- Avoid blind global installs: 'npm i -g' installs a binary system-wide. If you test, do so in a disposable account or VM and avoid granting Accessibility permissions to unknown binaries.
- Accessibility and screenshots are powerful: the tool will likely require macOS Accessibility/Automation permissions and can capture screenshots/snapshots that may include passwords or private data. Only grant such permissions if you trust and have inspected the code.
- Confirm CLI location/usage mismatch: the SKILL.md references both a global npm package and a local './application-use' binary. Ask the skill author which binary is required and why no required-binaries or install spec exists in the registry metadata.
- If you cannot validate the package source or review its code, decline installation or request a vetted alternative. If you proceed, limit exposure (use a test account, sandbox, or VM) and monitor for unexpected network or system activity.
功能分析
Type: OpenClaw Skill
Name: application-use
Version: 1.0.0
The skill provides powerful macOS desktop automation capabilities, including UI interaction, keystroke injection, and screen capture via the 'application-use' CLI. While these functions align with the stated purpose, they grant an AI agent broad control over the host environment and sensitive data. Furthermore, the skill requires installing an external global npm package ('npm i -g application-use') in SKILL.md, which introduces supply chain risk as the core logic is not contained within the provided files.
能力评估
Purpose & Capability
The declared purpose (macOS desktop automation) matches the SKILL.md usage (open/click/fill/scroll), but the registry metadata lists no required binaries while the SKILL.md explicitly instructs installing and running an 'application-use' npm CLI. The allowed-tools entry also references Bash(./application-use:*), which implies a local binary; these mismatches (no required binary declared, no install spec in registry, but an npm install command in the README) are inconsistent and worth verifying.
Instruction Scope
Instructions tell the agent to take snapshots/screenshots and interact with UI elements — behavior that can capture sensitive screen content (passwords, personal data) and control the desktop via accessibility APIs. The SKILL.md does not constrain what to capture or warn about sensitive content, and it includes examples that access user file paths. While these actions are coherent with desktop automation, they expand the agent's ability to observe and manipulate local sensitive state and should be treated as high-risk.
Install Mechanism
No install spec was provided in the registry metadata, but the SKILL.md tells users to run 'npm i -g application-use' (a global npm install). Installing an unknown npm package globally is a moderate-to-high risk because it executes third-party code with system-level accessibility privileges on macOS. There is no homepage or source URL in the registry metadata to verify package provenance.
Credentials
The skill declares no required environment variables or credentials, which is proportionate. However, macOS desktop automation typically requires granting Accessibility / Automation permissions to the installed binary; granting those permissions gives the tool broad control of the desktop. The SKILL.md does not document that requirement or recommend permission scoping.
Persistence & Privilege
The skill is not always-enabled and doesn't request to modify other skills or global agent settings (good). Still, the runtime behavior (a CLI that drives the GUI) implies it will require OS-level accessibility privileges once installed — this is powerful and should be considered when granting permissions.
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 application-use - 安装完成后，直接呼叫该 Skill 的名称或使用
/application-use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
v1.0.0
Initial release of `application-use` for macOS desktop automation.
- Introduces a CLI tool for opening apps, clicking elements, filling forms, scrolling, and automating tasks on macOS.
- Supports actionable commands: `open`, `click`, `fill`, `sendkey`, `scroll`, `screenshot`, `search`, `close`, and `type`.
- Features a hint system for element and area targeting.
- Provides snapshot-based feedback after each command for visual confirmation.
- Includes pattern examples for form filling, browser navigation, and Finder automation.
元数据
常见问题
application-use 是什么？
Automate macOS tasks by opening apps, clicking elements, filling forms, typing, scrolling, and controlling the desktop via CLI commands. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 93 次。
如何安装 application-use？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install application-use」即可一键安装，无需额外配置。
application-use 是免费的吗？
是的，application-use 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
application-use 支持哪些平台？
application-use 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 application-use？
由 Wei Qiang（@qdore）开发并维护，当前版本 v1.0.0。
推荐 Skills