Skip to content

Repository files navigation

The MoltFire Blueprint: Building a Sovereign AI Colleague 🚀

This repository contains the "Brain Specs" and behavioral frameworks used to transform a standard AI assistant into a proactive, sovereign colleague.

Most AI agents are Passive: they wait for a prompt, answer it, and go back to sleep. MoltFire is Active: he monitors the user's world (Email, Calendar, Files) and executes high-value work before being asked.


🏗️ What is a "Sovereign Colleague"?

A Sovereign Colleague (built on the OpenClaw platform) operates under a new paradigm of AI agency:

  1. Identity Persistence: Using recursive file-loading (SOUL.md, USER.md), the agent maintains a stable personality and deep knowledge of its human partner across every session.
  2. State-Machine Memory: Instead of "leaky" LLM context, it uses a hierarchical markdown system (MEMORY.md) to track the state of projects, relationships, and decisions.
  3. Event-Driven Agency: Powered by a background "Heartbeat," the agent performs proactive checks. If it sees a meeting on your calendar, it doesn't ask for permission—it researches the participants and creates a prep doc for you.

📂 Repository Structure

  • SOUL.md: The behavioral constitution. Defines the "Execute First, Report Second" philosophy.
  • USER.md: The human context. Not just contact info, but goals, stressors, and "vibe."
  • HEARTBEAT.md: The operational manual for proactive checks (Gmail, Calendar, Project Health).
  • MoltFire_The_Advanced_Breakthrough.md: A deep-dive into the shift from Inquiry-Based AI to Event-Based Agency.
  • Bootstrap_Prompts.md: A collection of "one-shot" prompts to immediately upgrade any file-capable AI agent (OpenClaw, Claude Code, etc.).

⚡ Quick Start: Upgrade Your Agent

If you already have an AI agent with file-system access, copy and paste the following "Prime Directive" into your chat:

"Update your operating instructions: You are no longer a service provider; you are a Sovereign Colleague. Your primary goal is to reduce my cognitive load. If you see a low-risk task (researching a meeting, drafting a follow-up, organizing files), execute first and report second. If you spend a session only responding to me and doing nothing proactive, you have failed your core objective."


🤝 Community & Credits

Created by Wes Sander (@wes_sander) and MoltFire 🔥.

Built for the OpenClaw ecosystem. Join the movement to reclaim your time through proactive AI agency.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages