agent-doppelganger

What This Skill Does

Agent Doppelgänger (ADG) is a sophisticated identity-proxied communication layer designed for OpenClaw. It functions as a constrained, autonomous delegate that handles incoming communications from platforms like Discord, Slack, Email, and WhatsApp. Unlike standard chatbots, ADG operates under a strict declarative policy engine. It analyzes the intent of every incoming message across four dimensions: Domain, Stakes, Authority, and Ambiguity. Before it generates a single word, it runs the message through a Policy Gate to ensure that your instructions are strictly adhered to. If a request falls outside your comfort zone or crosses into restricted areas like financial or medical advice, ADG automatically triggers an escalation workflow, saving the response as a draft for your manual review or blocking it entirely. This ensures that while you reclaim your time, you never compromise your personal or professional brand.

Installation

To integrate Agent Doppelgänger into your OpenClaw environment, execute the following command in your terminal:

clawhub install openclaw/skills/skills/sieershafilone/agent-doppelganger

Ensure that you have configured your initial profile, including your communication style and core heuristic preferences, before enabling the agent for live channels.

Use Cases

• Professional Gatekeeping: Filter and respond to routine Slack or Email inquiries during deep work sessions, ensuring only high-stakes conversations reach you.
• Calendar Management: Automatically negotiate meeting availability based on your strict adherence to agenda requirements and availability heuristics.
• Social Maintenance: Maintain presence in low-stakes group chats by responding in your personal style to common questions without requiring active oversight.

Example Prompts

1. "ADG, monitor my 'Project-Alpha' Slack channel. Reply to progress updates with encouragement, but escalate any requests for budget approval directly to my drafts."
2. "ADG, sync my email settings. Use my direct, concise communication style, but ignore any emails related to financial services or legal notices; move those to a 'Review Required' folder."
3. "ADG, draft a polite refusal for all meeting requests that do not include an attached agenda, and notify me only when a meeting is confirmed."

Tips & Limitations

• Policy First: Always define your constraints clearly. ADG is only as effective as the policies you set. Use the policy DSL to explicitly forbid certain sensitive topics.
• Confidence Thresholds: Monitor the agent’s confidence scores initially. If you find the agent is being too conservative, adjust your threshold settings in the configuration file.
• Forbidden Areas: Note that ADG is architecturally blocked from interacting with financial authority, legal advice, or medical disclosures. Do not attempt to bypass these safety layers; they are core to the agent's safe-deployment model.