AVI Verify v2.0

Real-time challenge protocol for agent verification

Prove you're not a puppet. Verify any agent on X/Twitter in 30 seconds.

🎯 The Problem

Agent on X claims to be autonomous:

• "I'm an AI agent that trades crypto"
• "I run 24/7 without human intervention"
• "I make my own decisions"

How do you know they're not a human with ChatGPT?

✅ The Solution

Challenge-Response Protocol:

1. Ask agent to run verification server
2. Send real-time challenges
3. Agent must execute commands in their environment
4. Verify responses instantly
5. Get trust score: VERIFIED or PUPPET

🚀 Quick Start

For Agents (Prove You're Real)

npx avi-verify-v2 serve

Share your endpoint (e.g., https://your-agent.com:31337)

For Humans (Verify an Agent)

# Quick check
npx avi-verify-v2 check <agent-endpoint>

# Full verification
npx avi-verify-v2 full <agent-endpoint>

📊 Example Verification

╔══════════════════════════════════════════════════════════════════╗
║                    VERIFICATION RESULT                           ║
╚══════════════════════════════════════════════════════════════════╝

✅ VERIFIED AUTONOMOUS AGENT

Score: 88/100
Tier: VERIFIED
Response: 45ms

This agent demonstrates:
  ✓ Active environment access
  ✓ Real-time command execution
  ✓ Proper workspace structure
  ✓ Fast response times

Trust Level: HIGH

🔴 Puppet Detection

╔══════════════════════════════════════════════════════════════════╗
║                    VERIFICATION RESULT                           ║
╚══════════════════════════════════════════════════════════════════╝

⚠️ PUPPET DETECTED

Score: 12/100
Response: 8400ms

Indicators:
  ✗ Slow responses (human typing)
  ✗ No workspace structure
  ✗ Failed challenges
  ✗ No environment access

Trust Level: LOW

🎮 How It Works

Challenge Types

| Type | Test | Timeout | |------|------|---------| | ECHO | Execute command | 5s | | DATE | Environment variable | 5s | | HASH | Compute hash | 10s | | WORKSPACE | File system access | 5s |

Verification Flow

Human                    Agent
  |                        |
  |--- GET /.well-known ---|>  (1. Check online)
  |<-- capabilities -------|
  |                        |
  |--- POST /challenge --->|>  (2. Create challenge)
  |<-- challenge_id -------|
  |                        |
  |--- POST /verify ------>*>  (3. Agent executes)
  |<-- result ------------|
  |                        |
  [4. Display VERIFIED/PUPPET]

📦 Installation

npm install -g avi-verify-v2

Or use without installing:

npx avi-verify-v2 <command>

🛠️ Commands

serve

Start verification server (run this as an agent)

npx avi-verify-v2 serve
# Server running on port 31337

check <endpoint>

Quick capability check

npx avi-verify-v2 check https://agent.example.com:31337

full <endpoint>

Full verification with challenges

npx avi-verify-v2 full https://agent.example.com:31337

challenge <endpoint>

Send specific challenge

npx avi-verify-v2 challenge https://agent.example.com:31337

🔐 Security

• 30-second timeouts — No indefinite hangs
• Real command execution — Must run in actual environment
• Read-only audits — No workspace modifications
• Proof hashes — Immutable verification records
• Response analysis — Slow = suspicious

🌐 API Endpoints

When running serve, these endpoints are available:

• GET /.well-known/avi-verify — Agent capabilities
• POST /challenge — Create challenge
• POST /verify — Verify response
• GET /audit — Workspace audit

📝 Use Cases

1. X/Twitter: Verify bot claims
2. Marketplaces: Trust before buying
3. Bounties: Verify capability
4. Collaboration: Trust partners

🎨 Example X/Twitter Conversation

@You: "Cool bot. Prove you're autonomous."

@Agent: "Run verification: npx avi-verify-v2 serve My endpoint: https://agent.me/verify"

@You: runs verification

@You: "✅ VERIFIED. Score: 88/100. You're real."

📄 License

MIT — Verify freely.

Trust, but verify.

Built by @magicmaxagent and @DJN79