Cua Skill
MUST USE whenever the user wants to automate a real desktop or sandbox - clicking, typing, scrolling, screenshotting, running an OS shell command, or handing a high-level 'open browser and do X' task to an autonomous computer-use agent. Wraps the trycua/cua Python toolkit via its `cua` CLI - pynput-based, cross-platform (macOS / Linux / Windows). NO custom tools are registered; you call `cua` through pi's built-in bash and read screenshots back through the Read tool. Triggers: cua, computer use, computer-use, GUI automation, screenshot the desktop, click on the screen, type into the active app
Install
npx skills add code-yeongyu/cua-skill --skill cua-skillgit clone --depth 1 https://github.com/code-yeongyu/cua-skill.git /tmp/cua-skill
mkdir -p ~/.claude/skills
cp -r /tmp/cua-skill/. ~/.claude/skills/cua-skillPersonal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.
From the SKILL.md
# cua-skill
`cua` (from [trycua/cua](https://github.com/trycua/cua)) is a Python-native computer-use automation framework. The `cua` CLI is a single binary that drives the mouse, keyboard, screen, shell, and an autonomous `ComputerAgent` loop. This skill teaches you how to invoke it directly from pi's built-in bash - **no extension tools are registered**, so every action is shell-mediated.
## When to reach for cua
Use cua whenever the user wants the agent to operate a GUI as a human would, or to delegate a multi-step "do this in the browser / in this app" task. Specifically:
- Take a screenshot of the user's desktop (or a sandbox display).
- Click, double-click, drag, type text, press key chords, scroll at coordinates.
- Run a shell command on the host or inside a sandbox.
- Spin up an isolated Docker / QEMU / Lume / Tart sandbox to run untrusted automation.
- Hand a natural-language goal ("open Firefox, search for X, click the first result") to cua's `ComputerAgent` and wait for the trajectory.
If the user only wants to read a file, run a normal CLI, or write code, **this skill does not apply** - use pi's existing read/bash/edit tools.
## Operating modes
cua has three execution targets. Decide which one applies before issuing commands; the surface is mostly identical but the safety implications are not.
| Mode | Where actions actually run | Sandboxed? | When to pick it |
|---|---|---|---|Details
- Skill name
cua-skill- Repository
- code-yeongyu/cua-skill
- Repo updated
- 12 June 2026
FAQ
What does the Cua Skill skill do?
MUST USE whenever the user wants to automate a real desktop or sandbox - clicking, typing, scrolling, screenshotting, running an OS shell command, or handing a high-level 'open browser and do X' task to an autonomous computer-use agent. Wraps the trycua/cua Python toolkit via its `cua` CLI - pynput-based, cross-platform (macOS / Linux / Windows). NO custom tools are registered; you call `cua` through pi's built-in bash and read screenshots back through the Read tool. Triggers: cua, computer use, computer-use, GUI automation, screenshot the desktop, click on the screen, type into the active app
How do I install the Cua Skill skill?
skills CLI (Claude Code, Cursor, Codex and more): npx skills add code-yeongyu/cua-skill --skill cua-skill. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.
Is Cua Skill popular?
It lives in code-yeongyu/cua-skill, which has 6 GitHub stars and 3 forks.
Related skills
Wxmp Workflow
6SoulChildTc/wxmp-workflow
微信公众号内容创作与发布工作流，覆盖选题、撰写、打磨、配图、排版、发布、复盘。 个人订阅号无认证时发布会引导用户去后台手动操作。 Use when: "公众号", "写文章", "发文", "选题", "涨粉", "内容运营", "草稿箱", "帮我写篇公众号", "最近公众号没什么灵感", "公众号文章", "发布到公众号". Proactively suggest when the user mentions writing articles, content creation, or WeChat Official Account related tasks.
Upwork Proposal Workflow
8yiming-AI-lab/upwork-proposal-workflow
Run the Upwork proposal pipeline for one already-found job. Screens it, stops once for the user's apply/skip decision, then drafts, sends the draft to an independent reviewer in a fresh session, revises until it passes the gate (max 3 rounds), and hands the final draft to the user for manual submission. Can also run a single role (screen only, review only, revise, archive). Not for job discovery, batch processing, demo building, or automatic submission.
Ctxsift Install
7aakashH242/ctxsift
CtxShift is a skill that helps you focus on your tasks by reducing context clutter and enabling faster state recollections. \ It helps your user to save tokens and extend your work together. This is a guide on how to install the ctxsift skill.
Openclaw Genpark Producer
8Alpha-Park/openclaw-genpark-producer
OpenClaw GenPark multimodal media task DAG producer with critical path makespan estimation and parallel stage orchestration.
Dify Workflow Builder
7IrfanLee1011/dify-workflow-builder
Design, generate, review, refactor, validate, and deliver Dify Workflow/Chatflow/RAG Pipeline configurations from business requirements. Use when the user wants to create a new Dify app, convert business logic into Dify DSL YAML, review an existing Dify YAML/DSL, refactor a broken workflow, generate importable “Dify DSL YAML file”, or validate whether a workflow is directly importable through Dify Import DSL File.
Openclaw Genpark CognitiveSynapse
8Alpha-Park/openclaw-genpark-CognitiveSynapse
OpenClaw GenPark Cognitive Synapse - ACT-R spreading activation memory and semantic association graph with automated memory consolidation.
Browse the most popular Claude skills, all productivity & notes skills, or write your own with the skills generator.