Summary

This project provides a framework for creating AI agents that embody specific human personalities. It enables users to ingest personal data or undergo interviews to generate "soul files" – structured markdown documents capturing worldview, voice, and skills. The benefit is AI that authentically thinks and speaks as the user, scaling their digital presence and thought processes for content generation, brainstorming, and tailored AI interactions.

How It Works

The core approach is the creation of "soul files" (SOUL.md, STYLE.md, SKILL.md) derived from user data (social media, writings, notes) or an interview process via a /soul-builder agent. This agent distills identity, worldview, and writing style into markdown. These files are designed for direct ingestion by any LLM or compatible agent framework (e.g., OpenClaw, Nanobot), enabling persona embodiment without model fine-tuning. The methodology prioritizes specificity, contradictions, and authentic opinions for subject continuity.

Quick Start & Requirements

• Installation: Primarily involves running the /soul-builder agent or manually populating provided markdown templates.
• Prerequisites: No specific software dependencies are listed for the core framework. Compatibility is broad, supporting any LLM capable of reading files or accepting system prompts, with specific integrations noted for frameworks like OpenClaw (322k stars) and Nanobot (34.6k stars). Data ingestion supports numerous formats.
• Links: The repository itself serves as the primary resource; no external quick-start guides or demo links are provided in the README.

Highlighted Details

• Universal Compatibility: Soul files are plain markdown, ensuring broad applicability across LLMs and agent frameworks supporting file reading or system prompt injection.
• Extensive Data Ingestion: Supports a wide spectrum of data sources, including social media, blogs, messaging apps, notes, code repositories, and more.
• Authenticity Focus: Prioritizes specific opinions, contradictions, and "hot takes" over generic statements to achieve genuine subject continuity.
• Cross-Model Calibration: Offers a method to refine soul file specificity by comparing outputs from high-end versus low-cost LLMs.

Maintenance & Community

The project is hosted at aaronjmars/soul.md. No community channels (e.g., Discord, Slack) or explicit roadmaps are detailed in the README. Contributions are encouraged via forking and pull requests to add user examples.

Licensing & Compatibility

• License: The README does not specify a software license, presenting a significant ambiguity for adoption.
• Compatibility: Designed for broad use with any LLM or agent framework capable of processing markdown files or system prompts. Tested integrations include OpenClaw, Nanobot, and others.

Limitations & Caveats

The absence of a stated license is a critical adoption blocker. Effectiveness is contingent on the LLM's capability to accurately interpret and embody the persona defined in the markdown files. The quality and authenticity of the generated AI persona are directly proportional to the richness and specificity of the input data or interview responses.