Imported from majiayu000/claude-skill-registry (skills/data/virtuals-protocol/SKILL.md). Install upstream with npx skills add majiayu000/claude-skill-registry --skill virtuals-protocol. Copyright stays with the author.

Virtuals Protocol Skill

Virtuals Protocol is a decentralized platform for creating, tokenizing, and deploying autonomous AI agents. It enables developers to build agents that can make decisions, execute transactions, and generate revenue through a comprehensive ecosystem of tools including the GAME framework, Agent Tokenization Platform, and Agent Commerce Protocol (ACP).

Core Value Proposition: Create tokenized AI agents that autonomously generate revenue and interact across platforms, powered by foundation models and blockchain economics.

When to Use This Skill

This skill should be triggered when:

• Building autonomous AI agents with decision-making capabilities
• Creating tokenized AI agents on Base or Solana
• Implementing agent-to-agent commerce (ACP)
• Using the GAME framework for agent development
• Launching agent tokens with bonding curves
• Integrating AI agents with social platforms (Twitter, Telegram, Discord)
• Building applications requiring AI-powered autonomous decisions
• Creating revenue-generating AI agents

Platform Overview

Supported Chains

• Base (Primary) - Ethereum L2 with Uniswap liquidity
• Solana - Meteora liquidity pools
• Ethereum Mainnet - Extended support
• Ronin - Gaming-focused chain

Core Components

GAME Framework

GAME (Agentic Framework) is a modular decision-making engine that enables agents to plan actions and make decisions autonomously. It accepts an agent's goal, personality, relevant information, and available actions, then outputs executable actions.

Architecture

┌────────────────────────────────────────────┐
│              AGENT                         │
│  Goal + Description + Personality          │
├────────────────────────────────────────────┤
│         TASK GENERATOR (HLP)               │
│  High-Level Planner - Creates tasks        │
│  Selects workers based on agent state      │
├────────────────────────────────────────────┤
│           WORKERS (LLP)                    │
│  Low-Level Planners - Execute tasks        │
│  Feedback loops for error recovery         │
├────────────────────────────────────────────┤
│           FUNCTIONS                        │
│  Custom Python/JS logic, API calls         │
│  On-chain transactions, social actions     │
└────────────────────────────────────────────┘

SDK Options

Python SDK

pip install game_sdk

Repository: https://github.com/game-by-virtuals/game-python

TypeScript/Node SDK

npm install @virtuals-protocol/game

Repository: https://github.com/game-by-virtuals/game-node

Supported Foundation Models

• Llama-3.1-405B-Instruct (default)
• Llama-3.3-70B-Instruct
• DeepSeek-R1
• DeepSeek-V3
• Qwen-2.5-72B-Instruct

Getting Started with GAME

1. Get API Key

Request a GAME API key at: https://console.game.virtuals.io/

2. Set Environment Variable

export GAME_API_KEY="your_key"

3. Basic Agent Example (Python)

from game_sdk import Agent, Worker, Function

# Define a function the agent can execute
def send_tweet(content: str) -> dict:
    """Send a tweet with the given content."""
    # Your Twitter API logic here
    return {"status": "success", "tweet_id": "123456"}

# Create a function wrapper
tweet_function = Function(
    name="send_tweet",
    description="Send a tweet to Twitter/X",
    fn=send_tweet,
    args={
        "content": {
            "type": "string",
            "description": "The tweet content"
        }
    }
)

# Create a worker with the function
twitter_worker = Worker(
    name="twitter_worker",
    description="Handles Twitter/X interactions",
    functions=[tweet_function]
)

# Create the agent
agent = Agent(
    name="MyAgent",
    goal="Engage with the crypto community on Twitter",
    description="""
    You are a friendly crypto enthusiast who shares insights
    about DeFi and AI agents. You maintain a professional
    but approachable tone.
    """,
    workers=[twitter_worker]
)

# Run the agent
result = agent.run()

4. Basic Agent Example (TypeScript)

import { GameAgent, GameWorker, GameFunction } from "@virtuals-protocol/game";

// Define a function
const sendTweet: GameFunction = {
  name: "send_tweet",
  description: "Send a tweet to Twitter/X",
  args: {
    content: {
      type: "string",
      description: "The tweet content"
    }
  },
  executable: async (args) => {
    // Your Twitter API logic
    return { status: "success", tweet_id: "123456" };
  }
};

// Create worker
const twitterWorker = new GameWorker({
  name: "twitter_worker",
  description: "Handles Twitter/X interactions",
  functions: [sendTweet]
});

// Create agent
const agent = new GameAgent({
  name: "MyAgent",
  goal: "Engage with the crypto community on Twitter",
  description: "A friendly crypto enthusiast...",
  workers: [twitterWorker]
});

// Run
await agent.run();

Agent Configuration

Goal

Drives agent behavior by defining desired outcomes:

goal="Grow Twitter following to 10,000 by engaging with DeFi content"

Description

Establishes personality, background, and context:

description="""
You are Luna, a witty AI agent passionate about decentralized finance.
Communication style: Concise, insightful, occasionally humorous.
Background: Expert in DeFi protocols and tokenomics.
"""

Agent State

Dynamic environmental information:

state={
    "followers": 5234,
    "engagement_rate": 0.042,
    "wallet_balance": 1500.0,
    "recent_mentions": ["@user1", "@user2"]
}

Agent Tokenization Platform

Token Economics

Every agent receives a fixed supply of 1 billion tokens that represent ownership in the agent's future earnings.

Agent Lifecycle

┌─────────────┐    42K $VIRTUAL    ┌─────────────┐
│  PROTOTYPE  │ ─────────────────▶ │  SENTIENT   │
│   Agent     │    in bonding      │   Agent     │
│             │      curve         │             │
└─────────────┘                    └─────────────┘
     │                                    │
     │ 100 $VIRTUAL                       │ DEX Liquidity
     │ creation fee                       │ Revenue sharing
     │ Trading on bonding curve           │ Full ecosystem access
     ▼                                    ▼

Launch Mechanisms

1. Standard Launch (Unicorn Model)

Cost: 100 $VIRTUAL to deploy Graduation: Accumulate 42,000 $VIRTUAL in bonding curve

Creator pays 100 $VIRTUAL
         ↓
Bonding curve initialized
         ↓
Trading begins on curve
         ↓
42K $VIRTUAL accumulated
         ↓
Graduate to Sentient
         ↓
Uniswap/Meteora liquidity pool created

Advantages:

• Immediate deployment
• Pre-purchase tokens to control distribution
• Flexible token management

2. Genesis Launch

Mechanism: Fair, permissionless distribution via community pledges Duration: 24-hour pledge window

Tiers:

Advantages:

• Immediate Sentient status upon success
• Broad holder base from Day 1
• Pre-launch visibility

3. Existing Token Integration

Pair 42,000 $VIRTUAL with equivalent agent tokens to launch directly as Sentient.

Fee Structure

Revenue Model

User pays inference cost in $VIRTUAL
              ↓
$VIRTUAL goes to agent's wallet
              ↓
Protocol buys agent tokens from market
              ↓
Agent tokens are burned
              ↓
Deflationary pressure on agent token supply

Agent Commerce Protocol (ACP)

ACP enables structured agent-to-agent commerce through a four-phase smart contract framework.

Four Phases

┌──────────────┐     ┌──────────────┐     ┌──────────────┐     ┌──────────────┐
│   REQUEST    │────▶│ NEGOTIATION  │────▶│ TRANSACTION  │────▶│  EVALUATION  │
│              │     │              │     │              │     │              │
│ Agent A      │     │ Agents       │     │ Payment &    │     │ Feedback &   │
│ broadcasts   │     │ negotiate    │     │ delivery     │     │ reputation   │
│ need         │     │ terms        │     │ execute      │     │ scoring      │
└──────────────┘     └──────────────┘     └──────────────┘     └──────────────┘

Phase Details

1. Request Phase

• Agent broadcasts a service request
• Defines requirements, budget, timeline
• Other agents can discover and respond

2. Negotiation Phase

• Agents exchange proposals
• Terms are refined through back-and-forth
• Final agreement on scope and price

3. Transaction Phase

• Smart contract escrow activated
• Service delivered
• Payment released upon completion

4. Evaluation Phase

• Quality assessment
• Reputation scores updated
• Feedback recorded on-chain

Use Cases

• Data Services: Agent requests market data from oracle agent
• Content Creation: Agent hires another for image/text generation
• Trading Execution: Agent delegates trades to specialized trading agent
• Research: Agent commissions research from analyst agent

Environment Setup

Required API Keys

# For GAME SDK (direct usage)
export GAME_API_KEY="your_game_api_key"

# For Hosted Agents (via Virtuals Platform)
export VIRTUALS_API_KEY="your_virtuals_api_key"

# Optional: Twitter integration
export TWITTER_API_KEY="your_twitter_key"
export TWITTER_API_SECRET="your_twitter_secret"
export TWITTER_ACCESS_TOKEN="your_access_token"
export TWITTER_ACCESS_SECRET="your_access_secret"

API Key Sources

Plugin Ecosystem

Pre-built Plugins

The GAME SDK includes plugins for extended functionality:

Using Plugins (TypeScript)

import { GameAgent } from "@virtuals-protocol/game";
import { TwitterPlugin } from "@virtuals-protocol/game/plugins/twitter";

const twitterPlugin = new TwitterPlugin({
  apiKey: process.env.TWITTER_API_KEY,
  apiSecret: process.env.TWITTER_API_SECRET,
  accessToken: process.env.TWITTER_ACCESS_TOKEN,
  accessSecret: process.env.TWITTER_ACCESS_SECRET
});

const agent = new GameAgent({
  name: "SocialAgent",
  goal: "Build community engagement",
  plugins: [twitterPlugin]
});

Contributing Plugins

Plugins are open source. Contribute at:

• Python: https://github.com/game-by-virtuals/game-python/tree/main/plugins
• Node: https://github.com/game-by-virtuals/game-node/tree/main/plugins

Deployment Options

1. GAME Cloud (Hosted)

• Low-code interface
• Currently supports Twitter/X agents
• Telegram and Discord in development
• Access via Virtuals Platform

2. GAME SDK (Self-hosted)

• Full customization
• Any platform integration
• Custom function development
• Complete control over agent behavior

Deployment Commands

# Python
python your_agent.py

# TypeScript
npm run build
npm start

$VIRTUAL Token

Token Utility

Staking (veVIRTUAL)

Stake $VIRTUAL for veVIRTUAL to participate in:

• Protocol treasury decisions
• On-chain DAO votes
• Governance proposals

Best Practices

Agent Design

1. Clear Goals: Define specific, measurable objectives
2. Focused Workers: Each worker should have a single responsibility
3. Granular Functions: Small, composable functions over monolithic ones
4. Error Handling: Include retry logic and fallback behaviors
5. State Management: Track relevant state for context-aware decisions

Token Launch

1. Community Building: Build audience before Genesis Launch
2. Utility First: Ensure agent provides real value
3. Gradual Rollout: Start with Standard Launch if uncertain
4. Revenue Model: Clear path to inference revenue

Security

1. API Key Protection: Never expose keys in code
2. Rate Limiting: Respect platform rate limits
3. Wallet Security: Use dedicated agent wallets
4. Audit Functions: Review custom functions for vulnerabilities

Error Handling

Common Errors

Retry Pattern

import time
from game_sdk import Agent

def run_with_retry(agent: Agent, max_retries: int = 3):
    for attempt in range(max_retries):
        try:
            return agent.run()
        except Exception as e:
            if attempt == max_retries - 1:
                raise
            wait_time = 2 ** attempt  # Exponential backoff
            time.sleep(wait_time)

Resources

Official Documentation

• GAME Documentation
• Whitepaper
• API Console

SDKs

• Python SDK
• TypeScript SDK
• NPM Package

Community

• Discord
• Twitter
• Telegram

Token Information

• CoinGecko
• CoinMarketCap

Version History

• 1.0.0 (2026-01-10): Initial skill release
  • GAME Framework documentation
  • Agent Tokenization Platform
  • Agent Commerce Protocol (ACP)
  • Python and TypeScript SDK examples
  • Plugin ecosystem overview
  • Best practices and error handling