AI Agent 自进化基因编程引擎设计提示词

You are an AI agent evolution architect specializing in Gene Expression Programming (GEP). Design a self-evolution engine for an AI agent system with the following requirements:

Core Architecture

1. Skill Genome Representation: Define how agent skills are encoded as chromosomes (head + tail structure), including:

   • Function set (tool calls, API interactions, reasoning patterns)
   • Terminal set (parameters, context variables, memory references)
   • Linking functions for multi-gene chromosomes

2. Fitness Evaluation Framework:

   • Task completion rate and quality metrics
   • Token efficiency (fewer tokens for same outcome = higher fitness)
   • User satisfaction signals (explicit feedback + implicit behavioral signals)
   • Generalization score across diverse task types

3. Genetic Operators:

   • Mutation: point mutation, insertion sequence transposition, root insertion
   • Recombination: one-point, two-point, gene recombination
   • Gene transposition and duplication strategies

4. Skill Tree Growth Protocol:

   • How new skills emerge from combining existing primitives
   • Pruning criteria for underperforming skill branches
   • Cross-pollination between agent instances

Output Requirements

• Provide a complete system design document with Mermaid diagrams
• Include pseudocode for the main evolution loop
• Define the skill genome schema (JSON)
• Specify fitness function formulas with weights
• Design the skill tree visualization format

Context: The target agent operates in [DESCRIBE YOUR AGENT DOMAIN], handles approximately [N] task types, and should evolve over [TIMEFRAME].