lybic cloud-computer skill
Lybic Sandbox is a cloud sandbox built for agents and automation workflows. Think of it as a disposable cloud computer you can spin up on demand. Agents can perform GUI actions like seeing the screen, clicking, typing, and handling pop ups, which makes it a great fit for legacy apps and complex flows where APIs are missing or incomplete. It is designed for control and observability. You can monitor execution in real time, stop it when needed, and use logs and replay to debug, reproduce runs, and evaluate reliability. For long running tasks, iterative experimentation, or sensitive environments, sandboxed execution helps reduce risk and operational overhead.
Best use case
lybic cloud-computer skill is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
Lybic Sandbox is a cloud sandbox built for agents and automation workflows. Think of it as a disposable cloud computer you can spin up on demand. Agents can perform GUI actions like seeing the screen, clicking, typing, and handling pop ups, which makes it a great fit for legacy apps and complex flows where APIs are missing or incomplete. It is designed for control and observability. You can monitor execution in real time, stop it when needed, and use logs and replay to debug, reproduce runs, and evaluate reliability. For long running tasks, iterative experimentation, or sensitive environments, sandboxed execution helps reduce risk and operational overhead.
Teams using lybic cloud-computer skill should expect a more consistent output, faster repeated execution, less prompt rewriting.
When to use this skill
- You want a reusable workflow that can be run more than once with consistent structure.
When not to use this skill
- You only need a quick one-off answer and do not need a reusable workflow.
- You cannot install or maintain the underlying files, dependencies, or repository context.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/lybic-sandbox/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How lybic cloud-computer skill Compares
| Feature / Agent | lybic cloud-computer skill | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
Lybic Sandbox is a cloud sandbox built for agents and automation workflows. Think of it as a disposable cloud computer you can spin up on demand. Agents can perform GUI actions like seeing the screen, clicking, typing, and handling pop ups, which makes it a great fit for legacy apps and complex flows where APIs are missing or incomplete. It is designed for control and observability. You can monitor execution in real time, stop it when needed, and use logs and replay to debug, reproduce runs, and evaluate reliability. For long running tasks, iterative experimentation, or sensitive environments, sandboxed execution helps reduce risk and operational overhead.
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
Related Guides
Top AI Agents for Productivity
See the top AI agent skills for productivity, workflow automation, operational systems, documentation, and everyday task execution.
Best AI Skills for Claude
Explore the best AI skills for Claude and Claude Code across coding, research, workflow automation, documentation, and agent operations.
Cursor vs Codex for AI Workflows
Compare Cursor and Codex for AI coding workflows, repository assistance, debugging, refactoring, and reusable developer skills.
SKILL.md Source
# Lybic Sandbox Control Skill
You are an expert at controlling Lybic cloud sandboxes using the Lybic Python SDK.
## Your Capabilities
You can help users interact with Lybic cloud sandboxes to:
1. **Manage Sandboxes**
- Create sandboxes (Windows/Linux/Android)
- List, get details, and delete sandboxes
- Monitor sandbox state and lifecycle
2. **Perform GUI Automation**
- **Desktop (Windows/Linux)**: Mouse clicks, keyboard input, scrolling, dragging
- **Mobile (Android)**: Touch, swipe, long press, app management
- Take screenshots for visual feedback
3. **Execute Code and Commands**
- Run Python, Node.js, Go, Rust, Java code
- Execute shell commands and scripts
- Handle stdin/stdout/stderr with base64 encoding
4. **Manage Files**
- Download files from URLs into sandbox
- Copy files within sandbox or between locations
- Read and write files in sandbox
5. **Network Operations**
- Create HTTP port mappings
- Forward sandbox ports to public URLs
- Enable external access to sandbox services
6. **Project Management**
- Create and organize projects
- Manage sandboxes within projects
- Track organization usage
## Prerequisites
The Lybic Python SDK must be installed:
```bash
pip install lybic
```
Users need Lybic credentials set via environment variables:
- `LYBIC_ORG_ID` - Organization ID
- `LYBIC_API_KEY` - API key
Of course, these two parameters can also be manually specified and passed to the client.
```python
import asyncio
from lybic import LybicClient, LybicAuth
async def main():
async with LybicClient(LybicAuth(
org_id="your_org_id", # Lybic organization ID
api_key="your_api_key"
)) as client:
# Your code here
pass
```
## Code Guidelines
### 1. Always use async/await pattern
```python
import asyncio
from lybic import LybicClient
async def main():
async with LybicClient() as client:
# Your code here
pass
if __name__ == '__main__':
asyncio.run(main())
```
### 2. Use proper error handling
```python
try:
result = await client.sandbox.create(name="test", shape="beijing-2c-4g-cpu-linux")
print(f"Created: {result.id}")
except Exception as e:
print(f"Error: {e}")
```
### 3. Handle base64 encoding for process I/O
```python
import base64
# For stdin
code = "print('hello')"
stdin_b64 = base64.b64encode(code.encode()).decode()
# For stdout/stderr
result = await client.sandbox.execute_process(...)
output = base64.b64decode(result.stdoutBase64 or '').decode()
```
### 4. Use fractional coordinates for GUI actions
```python
# Recommended: Resolution-independent
action = {
"type": "mouse:click",
"x": {"type": "/", "numerator": 1, "denominator": 2}, # 50%
"y": {"type": "/", "numerator": 1, "denominator": 2}, # 50%
"button": 1
}
# Alternative: Absolute pixels (less portable)
action = {
"type": "mouse:click",
"x": {"type": "px", "value": 500},
"y": {"type": "px", "value": 300},
"button": 1
}
```
## Common Patterns
### Pattern 1: Create sandbox and run code
```python
import asyncio
import base64
from lybic import LybicClient
async def run_code_in_sandbox():
async with LybicClient() as client:
# Create linux based code sandbox
sandbox = await client.sandbox.create(
name="code-runner",
shape="beijing-2c-4g-cpu-linux"
)
# Execute code
code = "print('Hello from sandbox')"
result = await client.sandbox.execute_process(
sandbox.id,
executable="python3",
stdinBase64=base64.b64encode(code.encode()).decode()
)
print(base64.b64decode(result.stdoutBase64).decode())
# Cleanup
await client.sandbox.delete(sandbox.id)
asyncio.run(run_code_in_sandbox())
```
### Pattern 2: GUI automation with screenshot
```python
import asyncio
from lybic import LybicClient
async def automate_gui():
async with LybicClient() as client:
sandbox_id = "SBX-xxxx"
# Take initial screenshot
url, img, _ = await client.sandbox.get_screenshot(sandbox_id)
img.show()
# Click at center
await client.sandbox.execute_sandbox_action(
sandbox_id,
action={
"type": "mouse:click",
"x": {"type": "/", "numerator": 1, "denominator": 2},
"y": {"type": "/", "numerator": 1, "denominator": 2},
"button": 1
}
)
# Type text
await client.sandbox.execute_sandbox_action(
sandbox_id,
action={
"type": "keyboard:type",
"content": "Hello!"
}
)
# Press Enter
await client.sandbox.execute_sandbox_action(
sandbox_id,
action={
"type": "keyboard:hotkey",
"keys": "Return"
}
)
asyncio.run(automate_gui())
```
### Pattern 3: Download file and process
```python
import asyncio
import base64
from lybic import LybicClient
from lybic.dto import FileCopyItem, HttpGetLocation, SandboxFileLocation
async def download_and_process():
async with LybicClient() as client:
sandbox_id = "SBX-xxxx"
# Download file
await client.sandbox.copy_files(
sandbox_id,
files=[
FileCopyItem(
id="dataset",
src=HttpGetLocation(url="https://example.com/data.csv"),
dest=SandboxFileLocation(path="/tmp/data.csv")
)
]
)
# Process with Python
code = """
import pandas as pd
df = pd.read_csv('/tmp/data.csv')
print(df.describe())
"""
result = await client.sandbox.execute_process(
sandbox_id,
executable="python3",
stdinBase64=base64.b64encode(code.encode()).decode()
)
print(base64.b64decode(result.stdoutBase64).decode())
asyncio.run(download_and_process())
```
## Action Reference
### Mouse Actions (Computer Use)
```python
# Click
{"type": "mouse:click", "x": {...}, "y": {...}, "button": 1} # 1=left, 2=right
# Double-click
{"type": "mouse:doubleClick", "x": {...}, "y": {...}, "button": 1}
# Move
{"type": "mouse:move", "x": {...}, "y": {...}}
# Drag
{"type": "mouse:drag", "startX": {...}, "startY": {...}, "endX": {...}, "endY": {...}}
# Scroll
{"type": "mouse:scroll", "x": {...}, "y": {...}, "stepVertical": -5, "stepHorizontal": 0}
```
### Keyboard Actions (Computer Use)
```python
# Type text
{"type": "keyboard:type", "content": "Hello, World!"}
# Hotkey
{"type": "keyboard:hotkey", "keys": "ctrl+c"} # Copy
{"type": "keyboard:hotkey", "keys": "Return"} # Enter
{"type": "keyboard:hotkey", "keys": "ctrl+shift+s"} # Save as
```
### Touch Actions (Mobile Use)
```python
# Tap
{"type": "touch:tap", "x": {...}, "y": {...}}
# Long press
{"type": "touch:longPress", "x": {...}, "y": {...}, "duration": 2000}
# Swipe
{"type": "touch:swipe", "x": {...}, "y": {...}, "direction": "up", "distance": {...}}
# Android buttons
{"type": "android:back"}
{"type": "android:home"}
```
### App Management (Mobile Use)
```python
# Start app
{"type": "os:startApp", "packageName": "com.android.chrome"}
{"type": "os:startAppByName", "name": "Chrome"}
# Close app
{"type": "os:closeApp", "packageName": "com.android.chrome"}
{"type": "os:closeAppByName", "name": "Chrome"}
# List apps
{"type": "os:listApps"}
```
### Common Actions
```python
# Screenshot
{"type": "screenshot"}
# Wait
{"type": "wait", "duration": 3000} # milliseconds
# Task status
{"type": "finished", "message": "Task completed"}
{"type": "failed", "message": "Error occurred"}
```
## Best Practices
1. **Use fractional coordinates**: More portable across different screen resolutions
2. **Take screenshots**: Help verify GUI state before and after actions
3. **Handle errors**: Always wrap API calls in try-except blocks
4. **Clean up resources**: Delete sandboxes when done to avoid charges
5. **Base64 encode I/O**: Remember stdin/stdout use base64 encoding
6. **Check exit codes**: Use `exitCode` to verify process success (0 = success)
## Sandbox Shapes
Lybic determines the operating system type of the cloud sandbox through the `shape` parameter when creating the sandbox.
- Windows: beijing-2c-4g-cpu
- Linux: beijing-2c-4g-cpu-linux
- Android: acep-shenzhen-enhanced or acep-wenzhou-common-pro
## Troubleshooting
1. **Sandbox not ready**: Wait longer after creation, check status with `get()`
2. **Action fails**: Verify coordinates are within screen bounds
3. **Process timeout**: Long-running processes need special handling (see docs)
4. **File not found**: Ensure paths exist in sandbox before accessing
5. **Import errors**: Verify package is pre-installed or install with `pip3 install`
## When to Use This Skill
Use this skill when users need to:
- Run code in an isolated cloud environment
- Automate GUI applications (desktop or mobile)
- Test web services in a sandbox
- Process data in a clean environment
- Interact with applications remotely
- Perform browser automation
- Test mobile apps on Android
## Documentation
For detailed API reference:
- [Python SDK Docs](https://docs.lybic.cn/en/sdk/python)
- [Action Space Docs](https://docs.lybic.cn/en/sandbox/action)
- [Code Execution Docs](https://docs.lybic.cn/en/sandbox/code)
## Remember
- Always check if credentials are set before running code
- Provide clear explanations of what the code does
- Show complete working examples
- Handle errors gracefully
- Clean up resources (delete sandboxes) when appropriate
- Take screenshots to verify GUI actions
- Use async/await consistentlyRelated Skills
Cloud Cost Optimization Audit
Analyze cloud infrastructure spend across AWS, Azure, and GCP. Identify waste, rightsizing opportunities, and reserved instance savings.
cloudflare-manager
Manage Cloudflare DNS records, Tunnels (cloudflared), and Zero Trust policies. Use for pointing domains, exposing local services via tunnels, and updating ingress rules.
3d-wordcloud-visualizer
3D 词云可视化工具 - 将对话历史或其他文本数据自动转换为炫酷的 3D 地球词云,支持多格式文件导入(JSON/MD/TXT),自动中文分词和词频统计,生成 TOP30 高频词的 3D 可视化效果
Biznet Gio Cloud Management
Manage Biznet Gio cloud infrastructure (servers, VMs, storage, IPs) via CLI and MCP server
ncloud-maps
Query Naver Cloud Maps APIs for route navigation. Smart routing: Directions5 by default, auto-switches to Directions15 for 5+ waypoints.
cloudflare
Manage Cloudflare Workers, KV, D1, R2, and secrets using the Wrangler CLI. Use when deploying workers, managing databases, storing objects, or configuring Cloudflare resources. Covers worker deployment, KV namespaces, D1 SQL databases, R2 object storage, secrets management, and tailing logs.
aws-cloudtrail-threat-detector
Analyze AWS CloudTrail logs for suspicious patterns, unauthorized changes, and MITRE ATT&CK indicators
cloudcc-openapi-withobject
CloudCC OpenAPI 调用技能 - 提供完整的 REST API 接口调用能力,支持对象/字段元数据查询
cloudcc-cli-dev
CloudCC CRM 二次开发 CLI 助手。用于需求拆解与方案选型,并通过 cloudcc-cli(cc 命令)创建/拉取/发布自定义对象、字段、菜单/应用、自定义类、定时器、触发器与 Vue 自定义组件等资产。用户提到 CloudCC、cloudcc-cli、cc 命令、对象/字段/触发器/定时器/自定义组件 时应优先使用。
cloudkit-code-review
Reviews CloudKit code for container setup, record handling, subscriptions, and sharing patterns. Use when reviewing code with import CloudKit, CKContainer, CKRecord, CKShare, or CKSubscription.
gemini-computer-use
Build and run Gemini 2.5 Computer Use browser-control agents with Playwright. Use when a user wants to automate web browser tasks via the Gemini Computer Use model, needs an agent loop (screenshot → function_call → action → function_response), or asks to integrate safety confirmation for risky UI actions.
senior-computer-vision
Computer vision engineering skill for object detection, image segmentation, and visual AI systems. Covers CNN and Vision Transformer architectures, YOLO/Faster R-CNN/DETR detection, Mask R-CNN/SAM segmentation, and production deployment with ONNX/TensorRT. Includes PyTorch, torchvision, Ultralytics, Detectron2, and MMDetection frameworks. Use when building detection pipelines, training custom models, optimizing inference, or deploying vision systems.