Awesome AI Agent Skills & Native MCP Servers 🚀

🌟 Why GenPark AI Agent Skills?

Every skill in this index is engineered according to the GenPark Production Standard:

• 🐍 Zero External pip Dependencies: 100% Python standard library (math, re, json, sys, time). Instant startup, zero package version hell.
• 🔌 Native Model Context Protocol (MCP): Plugs directly into Cursor IDE, Claude Desktop, Windsurf, and Smithery.ai.
• 🎯 100% Functional Parity: Deterministic mathematical models, spatial algorithms, and state machines instead of mocked placeholders.
• 🌍 Enterprise Portability: Ready to run in serverless AWS Lambda, Docker, Kubernetes sidecars, or local dev machines.

🏗️ Global Architecture

graph TD
    User([👨‍💻 Developer / User]) --> Client[💻 Cursor / Claude Desktop / Windsurf / Smithery]
    Client -->|JSON-RPC 2.0 via stdio| MCP[⚡ GenPark Native MCP Servers]
    
    subgraph Skill_Categories["📚 GenPark Skill Domains"]
        MCP --> Voice["🎙️ Real-Time Voice & Streaming Audio"]
        MCP --> Doc["📄 Document AI & Financial/Legal Extraction"]
        MCP --> Personal["👤 Personal Agent Life OS & Attention"]
        MCP --> Memory["🧠 Long-Term Memory & GraphRAG"]
        MCP --> Vision["👁️ Vision & Computer Use Grounding"]
        MCP --> Sec["🛡️ Cybersecurity & Guardrails"]
        MCP --> Swarm["🐝 Multi-Agent Swarm Orchestration"]
        MCP --> SRE["⚙️ Reliability & Context Deduplication"]
    end

📑 Categorized Skill Directory

🎙️ 1. Real-Time Voice Agent & Full-Duplex Audio (Phase 246)

Low-latency voice-to-voice agent infrastructure for OpenAI Realtime API, LiveKit, ElevenLabs, and Cartesia.

📄 2. Document AI & Complex Financial/Legal Extraction (Phase 247)

Extracting structured matrices, balance sheets, and cross-referenced contracts from unstructured documents.

👤 3. Personal AI Life OS & Executive Assistant (Phase 240 & 248)

Protecting human attention, building resilient habits, and minimizing decision fatigue.

🧠 4. Long-Term Memory & Continual Learning (Phase 245)

Stateful persistent memory, Ebbinghaus forgetting curves, and GraphRAG belief updating.

👁️ 5. Vision Agent & Computer-Use (Phase 244)

Grounding LLM reasoning into visual screen coordinates and DOM element trees.

🛡️ 6. Agentic Cybersecurity & Red-Teaming (Phase 243)

Protecting autonomous agents against prompt injection, unauthorized tool calls, and data leakage.

⚡ One-Click MCP Integration

1. Claude Desktop (claude_desktop_config.json)

{
  "mcpServers": {
    "voice-turn-taking": {
      "command": "python",
      "args": ["/path/to/genpark-voice-turn-taking-endpoint-detector-skill/mcp_server.py"]
    },
    "table-reconstructor": {
      "command": "python",
      "args": ["/path/to/genpark-hybrid-bounding-box-ocr-table-reconstructor-skill/mcp_server.py"]
    },
    "inbox-triage": {
      "command": "python",
      "args": ["/path/to/genpark-personal-inbox-cognitive-triage-skill/mcp_server.py"]
    }
  }
}

2. Cursor IDE Integration

Navigate to Cursor Settings > Features > MCP > Add New MCP Server:

• Type: command
• Command: python path/to/skill/mcp_server.py

3. Smithery.ai CLI

npx -y @smithery/cli install @Alpha-Park/genpark-voice-turn-taking-endpoint-detector-skill

🤝 Contributing & Ecosystem Submission

We welcome submissions of production-grade, zero-dependency AI Agent skills. Please see CONTRIBUTING.md for criteria:

1. 100% Python standard library only (no pip dependencies).
2. Native mcp_server.py JSON-RPC 2.0 implementation.
3. Test suite verifying deterministic execution.