AI导哈

AI智能体技能框架

Agent-Skill-Kit是一个本地优先、零API密钥的AI智能体技能框架，通过MCP、CLI和桌面GUI提供可插拔的技能管理与执行能力，适用于智能体开发、自动化工作流和技术情报获取。

  • by HughKnightOCE
  • Python
  • 本地部署
  • 1 星标
  • 更新于
在 GitHub 查看 HughKnightOCE/Agent-Skill-Kit-ASK-

使用说明

🤖 Agent-Skill-Kit (ASK) v2.0

Zero API Keys. Zero Signups. Zero Friction.
Agent-Native. MCP-Ready. Living & Breathing.
Local-first skills framework for AI agents. Clone. Run. Integrate.

🚀 Today's Agent Intelligence

Last Updated: October 28, 2026

📊 Top 5 Trending Tech Stories

#1. Ten days later, Git on macOS still trusts the compromised eMudhra certificates

#2. CNC Stone Carving

#3. Kentucky man becomes the first humanist granted religious exemption from AI

#4. Start Doing Graph Theory

#5. ICE Is Running One of the Largest Anti-Left Surveillance Ops in US History

This section is auto-updated daily with trending tech intelligence.

⚡ What is ASK?

Agent-Skill-Kit is a modern, AI-native framework that gives agents (Claude, Copilot, custom bots) local, keyless superpowers through the Model Context Protocol (MCP).

ASK provides self-contained skills that run on your machine using:

  • ✅ Local logic only (no cloud required)
  • ✅ FREE public APIs (no auth, no keys)
  • ✅ MCP Integration (native agent support)
  • ✅ Beautiful Live CLI (professional TUI)
  • ✅ YAML manifests (easy skill discovery)
  • ✅ Dynamic skill generation (The Architect)

Think of it as a "turbo-charged plugin system for AI agents—2026 edition."

⚡ Quick Start (30 seconds)

# Install from GitHub
pip install git+https://github.com/HughKnightOCE/Agent-Skill-Kit-ASK-.git

# View available skills  
ask dashboard

# Run a skill
ask run tech-pulse
ask run repo-visualizer
ask run architect "Your idea"

That's it! No .env files. No API key hunting. No OAuth redirects.

📥 Installation Options

Choose the option that fits your needs:

Option 1: pip Package - Recommended (Works everywhere)

Best for: All users (Windows, Mac, Linux)

# Install from GitHub
pip install git+https://github.com/HughKnightOCE/Agent-Skill-Kit-ASK-.git

# Or install from wheel file
pip install agent_skill_kit-2.0.0-py3-none-any.whl

# Then use the global command
ask dashboard
ask run tech-pulse
ask run repo-visualizer

Option 2: Source Installation - Best for Contributors

Best for: Developers, customization, contributing

git clone https://github.com/HughKnightOCE/Agent-Skill-Kit-ASK-.git
cd Agent-Skill-Kit-ASK-
pip install -e .

# Use the global command
ask dashboard
ask run architect "Convert CSV to JSON"

Option 3: Desktop GUI - Modern Interface

Best for: Visual users, no command-line needed

# Install with GUI support
pip install git+https://github.com/HughKnightOCE/Agent-Skill-Kit-ASK-.git

# Launch the beautiful desktop application
ask-gui

This opens a professional PyQt6 desktop application with:

  • 🎨 Modern, intuitive interface
  • 🔍 Skill browser and search
  • 📊 Real-time execution dashboard
  • 📜 Execution history tracking
  • ⚡ One-click skill execution with argument forms

Learn more about the GUI →

Option 4: Run Source Directly - No Installation

Best for: Quick testing, trying it out

git clone https://github.com/HughKnightOCE/Agent-Skill-Kit-ASK-.git
cd Agent-Skill-Kit-ASK-

# Run directly via CLI
ask dashboard
ask run tech-pulse
ask run architect "Your skill description"

# Or launch GUI (requires PyQt6)
pip install PyQt6
ask-gui

Option 5: Standalone Executable - Zero Dependencies

Best for: No Python installed, portable use

Download pre-built executables from Releases:

  • Windows: ask-gui.exe - Double-click to run
  • macOS: ask-gui - Run from terminal
  • Linux: ask-gui - Run from terminal

No Python installation required!

Windows Command Prompt Troubleshooting

If ask doesn't work in Command Prompt (cmd), use one of these alternatives:

Option A: Use PowerShell instead (Recommended)

ask dashboard
ask run tech-pulse
ask-gui  # GUI also works great in PowerShell

Option B: Use Python directly (Works everywhere)

python -m core.ask dashboard
python -m core.ask run tech-pulse

Option C: Add Scripts to PATH manually

setx PATH "%PATH%;%APPDATA%\Python\Python314\Scripts"

Then restart Command Prompt and try ask again.

📦 What's Included?

✨ V2.0 Features

🔌 MCP Server Integration

Skills are now exposed as native MCP tools via core/mcp_gateway.py. Claude and Copilot can discover and execute skills directly without custom integration.

# Claude can now see all skills as native tools
python core/mcp_gateway.py

🏗️ The Architect Skill

Generate new skills from natural language descriptions. The Architect creates skeleton Python scripts and YAML manifests instantly.

ask run architect "Convert CSV to JSON format"

💫 Living README

GitHub Actions automatically run Tech-Pulse every 24 hours and update this README's "Today's Agent Intelligence" section, proving the repo is alive and functional.

🎨 High-End Terminal UI

Enhanced TUI with:

  • ASCII art banner (ASK-2026)
  • System health dashboard
  • Live progress displays
  • Professional color-coded output

🖥️ Desktop GUI Application (NEW!)

Modern PyQt6 desktop interface featuring:

  • Skill Browser - Search and filter all available skills
  • Visual Executor - Execute skills with beautiful form inputs
  • Real-Time Output - Watch skill execution live
  • History Tracking - Review past skill executions
  • Dashboard - See system statistics and skill categories
  • Professional Design - Polished, native desktop application feel

Launch with: ask-gui

🎯 Four Zero-Config Skills

1. Repo-Visualizer 📊

Generates Mermaid.js class diagrams of your project structure instantly.

ask run repo-visualizer /path/to/project

Output: A clean Mermaid diagram ready to paste into documentation or feed to AI agents.

graph TD
    Root["MyProject"]
    Root --> skills["📁 skills"]
    Root --> core["📁 core"]
    skills --> visualizer["📁 repo-visualizer"]

2. Agent-Identity 🎭

Generates a unique SVG avatar for your project using the free DiceBear API.

ask run agent-identity "Agent-Skill-Kit"

Output: A beautiful, deterministic SVG avatar saved to disk and displayed.

<svg viewBox="0 0 100 100" xmlns="http://www.w3.org/2000/svg">
  <!-- Unique avatar based on project name -->
</svg>

3. Tech-Pulse 🔥

Fetches top 5 trending tech stories from HackerNews so your AI agent stays contextually aware.

ask run tech-pulse

Output: Markdown-formatted stories with links, scores, and comment counts.

#1. Breakthrough in AI Reasoning Model
- 🔗 URL: https://news.ycombinator.com/item/...
- 👤 Author: username
- 📈 Score: 1,234 | 💬 Comments: 456

4. The Architect 🏗️

NEW! Generate new skills from descriptions. The Architect creates starter templates automatically.

ask run architect "Analyze code complexity"

Output: A complete skill directory with script.py and manifest.yaml.

🎯 The Philosophy: Agent-Native + MCP-Ready

ASK is built specifically for AI agents. Each skill outputs:

  1. Clean Markdown (AI-readable format)
  2. Structured data (JSON, YAML in comments)
  3. Minimal noise (no interactive prompts)
  4. Error handling (graceful failures)

Perfect for:

  • 🤖 Claude Code - MCP integration for native skill access
  • 🧠 GitHub Copilot CLI - Custom MCP commands
  • 🔧 Custom agents - Extend with dynamic skill generation
  • 📡 LLM chains - Multi-step workflows without external APIs
  • 🏗️ AI frameworks - Drop-in MCP server for any agent

🔧 CLI Manager: ask (2026 Edition)

The high-performance TUI with ASK-2026 styling and live displays.

# Dashboard with system health (NEW!)
ask dashboard

# Run a skill with live progress
ask run <skill-name> [args...]

# System diagnostics
ask doctor

# Help
ask help

Features:

  • ✨ ASCII art banner (ASK-2026)
  • 📊 System health dashboard (Python version, Git status, dependencies)
  • 🎬 Live progress spinners
  • 🎨 Professional color-coded output

🌐 MCP Server Integration (NEW!)

Expose all skills as native MCP tools:

from core.mcp_gateway import SkillMCPServer

# Claude, Copilot, and other agents see skills as native tools
server = SkillMCPServer()
await server.run()

Claude can now:

User: "Use the repo-visualizer tool to analyze my project"
Claude: [invokes repo-visualizer via MCP] → sees complete output

🏗️ The Architect: Generate New Skills (NEW!)

Turn ideas into skills instantly:

# Create from natural language
ask run architect "Convert YAML to JSON"

# Output:
# ✅ Created: skills/convert-yaml-to-json/
# ✅ Created: script.py
# ✅ Created: manifest.yaml
# 🚀 Next: Edit and run your new skill!

💫 Living README (NEW!)

This README is automatically updated every 24 hours with trending tech intelligence via GitHub Actions. The "Today's Agent Intelligence" section proves ASK is:

  • ✅ Alive and functional
  • ✅ Running real workflows
  • ✅ Connected to real data
  • ✅ Production-grade

🔧 CLI Manager: ask.py

The Rust-inspired TUI powered by rich.

📁 Project Structure

Agent-Skill-Kit/
├── core/
│   └── ask.py                 # ⭐ Main CLI manager
├── skills/
│   ├── repo-visualizer/
│   │   ├── script.py          # Implementation
│   │   └── manifest.yaml      # Metadata
│   ├── agent-identity/
│   │   ├── script.py
│   │   └── manifest.yaml
│   └── tech-pulse/
│       ├── script.py
│       └── manifest.yaml
├── README.md                  # You are here
├── requirements.txt           # Dependencies
└── .github/
    └── workflows/             # CI/CD (optional)

🛠️ Dependencies

Minimal and lightweight:

PyYAML>=6.0        # For manifest parsing
rich>=12.0         # For beautiful CLI

Installation:

pip install -r requirements.txt

🤝 Contributing: Add Your Own Skill

ASK is designed for community contributions. Creating a skill takes 5 minutes.

Skill Template

Each skill needs:

  1. script.py - The logic
  2. manifest.yaml - Metadata

Step 1: Create the Skill Directory

mkdir -p skills/my-awesome-skill
cd skills/my-awesome-skill

Step 2: Write script.py

#!/usr/bin/env python3
"""My awesome skill that does X without any API key."""

def main(arg1, arg2=None):
    """Main execution function."""
    print("🎉 My skill output")
    print("✅ Ready for AI agents!")

if __name__ == "__main__":
    import sys
    arg1 = sys.argv[1] if len(sys.argv) > 1 else "default"
    main(arg1)

Step 3: Write manifest.yaml

name: My Awesome Skill
version: "1.0.0"
description: "Does X without requiring any API keys"

author: "Your Name"
category: "utility"

requirements:
  - name: "python"
    version: ">=3.8"

api_keys_required: false
dependencies: []

usage:
  command: "python script.py"
  args_required:
    - name: "arg1"
      description: "Description of argument"

output_format: "markdown"

tags:
  - "keyless"
  - "local"

Step 4: Test It

python ../core/ask.py dashboard
python ../core/ask.py run my-awesome-skill

Step 5: Submit a PR

We love contributions! Submit a PR with:

  • ✅ Working skill code
  • ✅ Proper manifest.yaml
  • ✅ Usage example in description
  • ✅ No external API keys required

🎁 Use Cases

For Individual Developers

# Morning ritual: Check tech news while coding
python ask.py run tech-pulse

# Visualize your codebase for documentation
python ask.py run repo-visualizer

For AI Agents (Claude Code, Copilot CLI)

# In your agent flow
subprocess.run(["python", "ask.py", "run", "repo-visualizer"])
# Parse the Mermaid output and feed to LLM

For Teams

# Onboarding: Generate project overview
python ask.py run repo-visualizer /monorepo/service-a

# Team branding: Create consistent avatar
python ask.py run agent-identity "TeamName"

🔐 Security & Privacy

✅ All processing is local - No data leaves your machine (except API calls)
✅ No authentication required - Zero credentials to manage
✅ Open source - Review the code, audit the logic
✅ Minimal dependencies - Less code = less risk

📊 Roadmap

  • [ ] More skills:
    • code-metrics: Analyze code complexity
    • dependency-graph: Visualize package dependencies
    • api-docs-generator: Auto-generate OpenAPI specs
  • [ ] Skill marketplace: Discover community skills
  • [ ] Skill templates: One-click skill generation
  • [ ] Agent integrations: Official Claude Code plugin
  • [ ] Performance: Caching layer for API calls

🐛 Troubleshooting

"ImportError: No module named 'rich'"

pip install rich pyyaml

"Skill not found"

ask dashboard  # Check installed skills

"API call failed"

Some free APIs have rate limits. Try again in a few seconds.

📄 License

MIT License - Free to use, modify, and distribute.

🌟 Show Your Support

  • ⭐ Star this repo if you find it useful
  • 🔗 Share it with other developers building AI tools
  • 📝 Contribute skills for the community
  • 💬 Open issues with feature requests

🤖 Made for Agents, By Developers

Agent-Skill-Kit is built with the belief that AI agents should have powerful, local tools without the friction of OAuth, API keys, and sign-ups.

Whether you're building a personal automation tool, a team assistant, or integrating with Claude Code, ASK is your zero-friction gateway to local intelligence.

Built with ❤️ for the AI-native future.

█████████████████████████████████████████████████████
█ Agent-Skill-Kit: Local Power. Zero Keys. Pure Joy. █
█████████████████████████████████████████████████████

相关 MCP