skills.rest
Mininglamp-AI
Mininglamp-AIOFFICIAL
Explore all skills by this author →
WebsiteGitHub

mano-cua

Automate desktop GUI tasks from natural language descriptions.

42|15|Updated Mar 26, 2026
Search Tags:#natural-language#desktop-automation#vision-model#local-inference#gui-automation#computer-use-agent#screen-interaction
One-click install
npx skills add https://github.com/Mininglamp-AI/mano-skill --skill mano-cua
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: mano-cua
Source: https://github.com/Mininglamp-AI/mano-skill/tree/main/skill
Command: npx skills add https://github.com/Mininglamp-AI/mano-skill --skill mano-cua

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Manually performing repetitive or complex desktop GUI tasks across apps and websites with no native API or CLI support is tedious and time-consuming. This Skill eliminates that manual work by turning natural language task descriptions into automated screen interactions.

Core Features & Use Cases

  • Hybrid Vision Model Routing: Automatically selects the lightweight Mano-P model for simple repetitive tasks or the Claude CUA model for complex multi-step reasoning tasks to balance speed and accuracy.
  • Dual Inference Modes: Supports cloud mode for broad compatibility and fully on-device local mode (macOS Apple Silicon only) for privacy-sensitive tasks with zero data leaving the machine.
  • Safety Guardrails: Includes step limits, confirmation prompts for irreversible actions, and real-time status tracking to prevent unintended system changes.
  • Use Case: For example, you can use it to fill out web registration forms, navigate to specific settings in desktop apps, create and tag files in your file manager, or perform multi-step browser research tasks without writing any custom automation scripts.

Quick Start

Use the mano-cua skill to search for the latest AI news on Google and open the first search result.

Frequently Asked Questions about mano-cua

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate desktop GUI tasks that lack CLI tools or native APIs?▼

You can automate desktop GUI tasks without native APIs by using natural language descriptions to trigger visual screen interactions. This approach interprets your instructions to perform cross-platform actions like app navigation and form filling automatically.

Can I run local inference for GUI automation on macOS Apple Silicon?▼

Yes, local inference for GUI automation is supported on macOS Apple Silicon devices. This on-device mode ensures privacy-sensitive tasks are processed locally with zero data leaving your machine.

How does vision model routing work for multi-step screen interaction tasks?▼

Vision model routing for screen interaction automatically selects a lightweight model for simple repetitive tasks or a complex reasoning model for multi-step workflows. This hybrid routing balances processing speed and task accuracy.

What safety guardrails exist for preventing unintended system changes during automated form filling?▼

Safety guardrails for automated form filling include enforced step limits, real-time status tracking, and confirmation prompts for irreversible actions. These precautions prevent unintended system changes during execution.

Does cross-platform desktop automation work the same on Windows and Linux?▼

Cross-platform desktop automation supports web browsing, app navigation, and file management uniformly across Windows, Linux, and macOS. Broad compatibility is maintained through cloud inference mode for these operating systems.

Do I need the mano-cua binary to perform natural language desktop automation?▼

Yes, the mano-cua binary version 1.1.0 or higher is required to perform natural language desktop automation. This local binary executes the visual interactions and hybrid model routing on your system.

Related Skills

Explore More Software Engineering Skills→
PicoMLXPicoMLX
24

swift-mlx-lm

Run local LLM and VLM inference on Apple Silicon using MLX in Swift.

Official
Advanced
aggelosChatziioannouaggelosChatziioannou

Transparent Thinking

Send real-time status updates on AI task progress.

Community
Intermediate
AradotsoAradotso
2

photo-agents-autonomous-llm

Create autonomous vision-grounded agents with layered memory and CDP browser automation.

Official
Advanced
ClaudioCeppi83ClaudioCeppi83

nano-pdf

Edit PDF text on specific pages using natural language commands.

Community
Basic