Skip to content

About

GenPark & OpenClaw Premium Growth Hub, Showcase and Live Skill Generator.

Topics

Resources

Stars

8 stars

Watchers

0 watching

Forks

Latest commit

 

History

9 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

GenPark & OpenClaw 🦞

The Private Personal AI Assistant Ecosystem That Actually Does Things.
Native multi-messenger integration (WhatsApp, Telegram, Slack, iMessage), resident browser computer-use, and modular skills platform.

License: MIT OpenClaw Gateway Skills Registry PRs Welcome GitHub Stars


🌟 What is GenPark?

GenPark is an open-source, local-first personal AI assistant framework powered by the OpenClaw gateway daemon. While conventional chatbots are trapped in web windows answering questions, GenPark agents execute complex, multi-step tasks across your physical and digital world:

  • 🔒 Local-First & Private: Your credentials, cookies, and chat history remain on your machine inside the OS Secure Enclave / Local Vault. No cloud vendor data harvesting.
  • 💬 Zero-New-Interface: Coordinate everything natively through the apps you already live in—WhatsApp, Telegram, Slack, and iMessage.
  • 🌐 Resident Browser Execution: Solves the bot-blocking and IP-banning issues that plague commercial cloud agents by leveraging your resident browser profile and residential IP.
  • 🤝 Agent-to-Agent (A2A) Mesh: Peer agents coordinate shared shopping, group discounts, and split billing via encrypted zero-knowledge handshakes.
  • 🛡️ Active Detective Shield: Runtime pre-execution interception that prevents model hallucinations and stops PII address leaks before network dispatch.
  • ⚡ Personal Agent Protocol (PAP) Bridge: Standardized OAuth merchant interactions with fallback to resident browser automation for closed platforms.
  • 🎯 0% Ad Curation & In-Chat Checkout: Anti-sponsorship neutral recommendations and instant headless purchasing.
  • 💡 M2AI Merchant Infrastructure: Transforms online stores into agent-ready platforms via Agentic Pages, GEO/AEO optimization, and agent GMV attribution.
  • 🎨 Scene-Aware Aesthetic Styling (Look-to-SKU): Bridges real-life occasions and vibes with virtual styling and instant SKU procurement.
  • 🌐 Multi-Rail Agent Distribution: Serves context cards across ACP/UCP/MCP rails while keeping checkout sovereign on the brand's store.
  • 🔄 Closed-Loop Dynamic Funnel Optimization: Replaces fragmented SaaS stacks with intent-driven dynamic landing generation.
  • 👁️ Sensory Micro-Telemetry & In-Scroll Cards: Reads trackpad/touchscreen linger dynamics on PDPs to sequence narrative cards and in-scroll search.
  • 🧩 Extensible Skills Architecture: Plug-and-play modular skills defined in standard SKILL.md format.

🔬 Featured Research: Autonomous Decision Agent Architecture & Distillation

We conducted an exhaustive architectural deconstruction analyzing the state-of-the-art 2026 personal agent paradigms:

  • Cloud-Sentinel VM Paradigm: Cloud-isolated sandboxes with permission brokers and tokenized checkout.
  • Reflex-Concierge Paradigm: Verbal reinforcement learning (Reflexion) loops with zero-new-interface multi-channel messaging and concierge calling.
  • Trusted Agent Mesh & Active Detective (Late-Sept 2026 Updates): Direct Agent-to-Agent (A2A) coordination and runtime hallucination/PII guardrails.
  • Personal Agent Protocol & Dedicated Procurement (October 2026 Advances): Open PAP handshake standard (/.well-known/poppy.json), anti-ad procurement (Marc Lore / Wizard model), and buyer vs. seller AI dynamics.
  • M2AI & Merchant-Side Infrastructure (DeepLumen Distillation): Transitioning web stores from human visual rendering to machine-readable Agentic Pages, GEO/AEO citations, and agentic attribution.
  • Aesthetics & Scene-Aware Inspiration-to-Shopping (ViBA Distillation): Transitioning from rigid spec comparison to vibe-driven styling, Angel/Devil dual persona, and Look-to-SKU seamless commerce.
  • Multi-Rail Agent Distribution & Sovereign Checkout (Nile Distillation): Context cards as the new shelf, broadcasting across ACP/UCP/MCP/Skills, and sovereign checkout routing.
  • AI-Native Conversion & Closed-Loop Funnels (FERMÀT Distillation): Replacing fragmented e-commerce SaaS stacks with intent-driven dynamic landing generation and revenue recovery.
  • Sensory Micro-Telemetry & PDP Seller Agent (Product Genius Distillation): "Sensitive skin of a robot" touch/scroll parsing, dynamic in-scroll story cards, in-scroll search, and ad-platform intent feedback.

📖 Read the full architectural research report:
👉 自主个人决策与代购 Agent 架构解构与能力蒸馏

Key Architectural Distillations in GenPark:

  1. Local-First Security Enclave: Replaced cloud VM credential custody with a local hardware-backed vault.
  2. Reflexion-Driven Decision Engine: Integrated self-reflective trial-and-error loops for autonomous shopping and web navigation.
  3. Anti-Bot Resident Session Relay: Bypasses merchant anti-bot protections by operating within the user's authentic local browser context.
  4. Sentinel Human-In-The-Loop (HITL) Gate: Pre-stages checkout actions and sends one-tap interactive approval cards directly to your WhatsApp or Telegram before committing payments.
  5. A2A Trusted Negotiation Mesh: Decentralized P2P agent protocol for collaborative bulk purchases and zero-knowledge calendar alignment.
  6. Active Detective In-Flight Shield: Intercepts ungrounded parameter hallucinations and prevents accidental address leaks during form submission.
  7. PAP Hybrid Bridge & Adaptive Routing: Connects via OAuth Poppy endpoints where available, and falls back to local browser emulation on Amazon/walled gardens.
  8. Anti-Ad Curated Triad: Delivers exactly 3 uncompromising choices with 0% sponsor bias and one-tap in-chat checkout.
  9. M2AI & Agentic Pages Merchant Stack: Empowers merchants to serve machine-readable shadow catalogs, inject GEO/AEO entity graphs, and attribute agent GMV.
  10. Scene-Aware Vibe & Look-to-SKU Loop: Translates real-life occasions and aesthetics into verified merchant SKUs with dual-persona styling advice.
  11. Multi-Rail Context Broadcasting: Serves lightweight ACP, UCP, MCP, and Agent Skills manifests while keeping checkout sovereign on the brand's store.
  12. Closed-Loop Funnel Adaptation: Transforms raw session behavior into dynamic, intent-aligned landing experiences without static SaaS bloat.
  13. Real-Time PDP Seller Engine: Interleaves algorithmic story cards and semantic search inside natural scroll streams to engage silent shoppers.

🛠️ Flagship Skills: The Decision, Procurement & Merchant Suite

Based on our architectural research, we officially released the Agentic Commerce Ecosystem:

Skill Module Breakthrough Distillation
🛒 zenchoice-ai-skill ZenChoice (禅选) Autonomous multi-site shopping, Reddit sentiment audit, MAUT trade-off scoring, and staged cart checkout.
🎯 zenprocure-ai-skill ZenProcure (禅购) Dedicated 0% ad curation, Curated Triad paralysis-breaker, and universal headless in-chat checkout.
🎨 zenviba-ai-skill ZenViba (禅韵) Inspiration-to-Shopping agent: Scene-aware aesthetic styling, Angel & Devil dual-persona critique, and Look-to-SKU procurement.
💡 zenlumen-ai-skill ZenLumen (禅明) M2AI & Agentic Commerce merchant infrastructure: Agentic Pages, GEO/AEO optimization, and agent attribution.
🌊 zennile-ai-skill ZenNile (禅川) Multi-rail agent distribution (ACP/UCP/MCP/Skills), Context Cards compiler, and sovereign checkout handoff.
⚡ zenfermat-ai-skill ZenFermat (禅策) Closed-loop funnel adaptation, behavior-to-intent analysis, dynamic experience synthesis, and SaaS tool consolidation.
🧠 zengenius-ai-skill ZenGenius (禅灵) Real-time PDP seller agent, sensory micro-gesture telemetry ("sensitive skin"), in-scroll story cards, and ad intent feedback.
🌉 zenbridge-ai-skill ZenBridge (桥联) Hybrid Personal Agent Protocol (PAP) OAuth client + Resident Browser adaptive fallback.
🤝 zenpeer-ai-skill ZenPeer (跨体协行) Encrypted Agent-to-Agent (A2A) peer negotiation for group buying arbitrage, zero-knowledge bookings, and split billing.
🛡️ aegis-guard-skill Aegis Guard (行前护盾) Active Detective runtime anti-hallucination interceptor and PII address/credential redaction firewall.

📦 Skills Registry Catalog

GenPark comes bundled with production-grade skills:

Skill Category Description
🛒 zenchoice-ai-skill Personal Agent / Decisions ZenChoice autonomous shopping, multi-site comparison, Reddit sentiment audit, and staged checkout.
🎯 zenprocure-ai-skill Commerce / Procurement Dedicated 0% ad curation, Triad recommendation, and universal in-chat checkout.
🎨 zenviba-ai-skill Commerce / Aesthetic & Lifestyle Distilled from ViBA. Scene-to-vibe styling room, Angel/Devil dual persona, and Look-to-SKU in-chat procurement.
💡 zenlumen-ai-skill Commerce / Merchant M2AI Distilled from DeepLumen. Agentic Pages compiler, GEO/AEO semantic optimizer, and agent traffic attribution.
🌊 zennile-ai-skill Commerce / Agent Distribution Distilled from Nile. Multi-rail context cards (ACP/UCP/MCP/Skills), agent salesperson, and sovereign merchant checkout.
⚡ zenfermat-ai-skill Commerce / Funnel Optimization Distilled from FERMÀT. Behavior-to-intent analysis, dynamic landing synthesis, and SaaS consolidation.
🧠 zengenius-ai-skill Commerce / PDP Personalization Distilled from Product Genius. Micro-gesture tactile sensor, in-scroll story cards, in-scroll search, and ad intent telemetry.
🌉 zenbridge-ai-skill Commerce / Protocol Personal Agent Protocol (PAP) OAuth handshake with resident browser fallback.
🤝 zenpeer-ai-skill Personal Agent / Collaboration Agent-to-Agent (A2A) peer negotiation, collective buying arbitrage, and privacy-preserving bookings.
🛡️ aegis-guard-skill Security & Guardrails Pre-execution active hallucination interceptor and token-level PII address/card leakage shield.
💳 swap-commerce-skill Commerce Autonomous commerce coordinator for catalog sync, cross-border tax, and returns routing.
🔍 trooly-ai-skill Analytics Qualitative customer empathy engine and review scraping blueprint builder.
🚀 mktflywheel-ai-skill Marketing Autonomous CMO agent loop for cross-channel campaigns, ad copy, and SEO.
📢 nexad-ai-skill Marketing Multi-variant creative copywriting and visual hierarchy asset testing.
🎭 character-ai-skill Simulation Sandboxed character customization and customer-support agent simulator.
📈 tradingagents-skill Finance Algorithmic trading harness and local backtesting execution loop.

🚀 Quick Start

1. Initialize OpenClaw Personal Gateway

# Clone the repository
git clone https://github.com/Alpha-Park/genpark-hub.git
cd genpark-hub

# Start the local gateway daemon
openclaw gateway --start
# ✔ Gateway bound to 127.0.0.1:18789 (Launchd/systemd active)
# ✔ Paired Messengers: WhatsApp [active], Telegram [active]

2. Invoke the ZenChoice Agent via Messenger

Simply text your paired WhatsApp or Telegram bot:

"Find me over-ear noise-canceling headphones under $250 for long flights. 
Compare reviews on Reddit, check Best Buy vs. Amazon, and stage the best deal in my cart."

The agent will:

  1. Conduct real-time research in a sandboxed resident browser.
  2. Return a structured comparative decision matrix with pros, cons, and net landed costs.
  3. Stage the best item in your cart and send an interactive approval card with a [Confirm Purchase] button.

🌐 Web UI & Skill Builder

GenPark includes a Kyoto-minimalist interactive dashboard and visual SKILL.md builder. Simply open index.html in your browser or serve it via:

npx serve .
# Visit http://localhost:3000

🤝 Contributing & Community

Contributions are welcomed! Feel free to submit custom skills, enhance the OpenClaw gateway daemon, or improve our decision models.

  1. Fork the Project
  2. Create your Feature Branch (git checkout -b feature/NewSkill)
  3. Commit your Changes (git commit -m 'Add NewSkill')
  4. Push to the Branch (git push origin feature/NewSkill)
  5. Open a Pull Request

⭐ Support Us

If you believe in private, autonomous personal agents that work for individuals rather than big tech platforms, please give us a Star on GitHub!

Star on GitHub

About

GenPark & OpenClaw Premium Growth Hub, Showcase and Live Skill Generator.

Topics

Resources

Stars

8 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages