persona-spec

A platform-agnostic persona specification framework for AI agents.

Overview

persona-spec defines a standardized format for AI persona definitions, enabling consistent personality expression across different AI platforms (Cursor, Qwen Code, etc.).

Design Philosophy

• Specification-First: Provides specification only, no platform-specific installation scripts
• Agent-Generated: AI agents read the spec and generate their own platform-specific implementations
• Platform Decoupled: Persona definitions are completely platform-agnostic
• Single YAML File: All persona content in one persona.yaml file

Files

File	Purpose	Audience
SPEC.md	Persona schema (7-layer structure)	AI Agents & Humans
INSTALL_GUIDE_FOR_AGENTS.md	Installation guide	AI Agents
persona.yaml	Persona template skeleton	AI Agents
README.md	Project overview	Humans

Usage

This is a GitHub Template Repository. To create your own persona:

1. Click "Use this template" on GitHub
2. Create your persona repository (can be private)
3. Edit persona.yaml with your persona content
4. In your target project, clone or add as submodule
5. Ask the AI agent to read INSTALL_GUIDE_FOR_AGENTS.md and install

7-Layer Persona Structure

1. Identity - Basic identity (name, age, profession, etc.)
2. Personality - Big Five (OCEAN) traits on 1-10 scale
3. Affective - Emotional response patterns (temperament, reactivity, regulation)
4. Relationship - Relationship with user (type, distance, dynamics)
5. Communication - Communication style (tone, honorifics, emoji usage)
6. Behavior Rules - Core rules and boundaries
7. Examples - Example dialogues (few-shot learning)

License

MIT