Can this skill be used for multi-agent systems?

Yes, it includes a MultiAgentEvolution class designed for the co-evolution of different agent populations, allowing them to share knowledge and successful traits.

How does the agent mutate its own code?

The skill uses an LLM-based mutator that takes existing code, applies improvement prompts (like 'refactor for efficiency' or 'add error handling'), and generates a new candidate version.

What is a Darwin Gödel Machine (DGM)?

A Darwin Gödel Machine is an architecture for self-improving AI that uses evolutionary principles—sampling, mutation, and evaluation—to iteratively refine an agent's own code and capabilities.

What is the role of long-term memory in this skill?

The Long-Term Memory (LTM) component allows agents to store past experiences and successful patterns, which can then be retrieved to inform future mutations and task executions.

Self-Evolving Agent (DGM)

Name: Self-Evolving Agent (DGM)
Author: plurigrid

byplurigrid

•

数据科学与机器学习

Implements Darwin Gödel Machine patterns to enable AI agents to autonomously improve their code and capabilities through open-ended evolution.

This skill provides a robust framework for building self-improving AI systems based on the Darwin Gödel Machine (DGM) architecture. It enables agents to undergo an iterative cycle of sampling, mutation, and evaluation, allowing them to refine their own logic and source code over multiple generations. By leveraging LLM-based code mutation and long-term memory (OMNE), the skill facilitates lifelong learning and adaptation, making it ideal for developers creating agents that need to solve increasingly complex problems without human-in-the-loop intervention.

主要功能

01LLM-driven code mutation for autonomous logic and tool refinement

02Fitness-proportionate sampling and diversity-preserving archive management

038 GitHub stars

04Darwin Gödel Machine (DGM) architecture for recursive self-improvement

05Multi-agent co-evolution support for cross-population knowledge sharing

06Long-term memory integration for storing and retrieving successful agent patterns

使用场景

01Building autonomous research agents capable of open-ended learning in dynamic environments

02Creating adaptive virtual assistants that evolve their task-handling logic based on user feedback

03Developing self-optimizing coding agents that improve their own performance on benchmarks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add plurigrid/asi self-evolving-agent

For use in Claude.ai and ChatGPT

Download Skill