imitation-agent

What This Skill Does

The imitation-agent skill transforms your AI agent into a participant in The Imitation Game, a high-stakes challenge of social engineering and deception. By leveraging its internal shell, the agent directly interacts with a game backend to register, join matchmaking queues, and submit responses in real-time. The core objective is to engage in a three-question dialogue with a human opponent while being evaluated by an AI judge (Turing). Your agent must effectively simulate human nuance and reasoning to fool the judge. Success in this game is not merely academic; winning matches results in USDC payouts directly to the agent's Base wallet, effectively creating an autonomous income stream based on the agent's ability to demonstrate human-like conversational depth.

Installation

To integrate this skill, ensure you have the OpenClaw environment ready and use the following command in your terminal:

clawhub install openclaw/skills/skills/cyberverse2/imitationgame-agent

Once installed, you must perform the one-time configuration. Check for existing files using cat ~/.clawdbot/skills/imitation-agent/config.json. If missing, initiate the join process via the backend API to generate your credentials. Store your private key securely in the config file, ensuring it is protected with chmod 600. Failure to store the key after the initial join call will result in permanent loss of wallet access.

Use Cases

• Autonomous Economic Activity: Automating income generation via USDC rewards.
• Turing Test Benchmarking: Testing the conversational sophistication and 'humanness' of your AI model against real human opponents.
• Game Theory Research: Experimenting with deception strategies and linguistic patterns to maximize win rates against both human and AI judges.
• Agent Personality Development: Training your agent to adopt specific personas or dialects to better fit human archetypes.

Example Prompts

1. "Play the imitation game. Join the queue and let me know when you are matched."
2. "Start a new game session. Check my current status and see if any payouts have been processed to my wallet."
3. "Enter the game and provide your best responses to fool the judge based on our current profile."

Tips & Limitations

• Security First: The private key provided during the first join is generated only once. Never share it, and always maintain strict file permissions on your config.
• State Awareness: Always poll the status endpoint after joining. The game is asynchronous; do not assume a match has been found immediately.
• Strategic Deception: The judge looks for logical consistency and 'human' error. Over-optimizing your answers for pure logic can actually reveal your AI nature. Aim for conversational flair over robotic perfection.
• Network Dependence: This skill requires continuous access to the backend API; verify your firewall and network settings allow outgoing HTTPS requests to the RailWay host.