What does V-Evolve actually do?

V-Evolve acts as an evolution engine for Claude Code; it detects when Claude lacks a specific tool or agent to complete a task and builds the necessary capability on the fly.

Does V-Evolve learn from coding mistakes?

Yes, it maintains a `lessons-learned.md` file where it logs task failures, root causes, and prevention strategies to ensure it doesn't repeat the same errors.

How does V-Evolve decide which AI model to use for new agents?

It follows a complexity guide: Haiku is used for fast/simple tasks, Sonnet for balanced workloads, and Opus for complex reasoning and critical decisions.

Where are the evolved skills and agents stored?

All new creations are stored locally in your `~/.claude/agents/` and `~/.claude/skills/` directories, making them persistent across different sessions.

Is the evolution process automatic or manual?

It is semi-autonomous; the Proactive Evolution Protocol identifies opportunities and proposes a solution, which it then executes unless the user objects.

V-Evolve: Self-Improving Agent Engine

Name: V-Evolve: Self-Improving Agent Engine
Author: kks0488

bykks0488

•

生产力与工作流

Automates the creation of specialized agents and skills to expand Claude Code's native capabilities through a continuous self-evolution loop.

V-Evolve is a meta-skill designed to overcome the static limitations of AI assistants by implementing a self-improvement engine within the Claude Code environment. It monitors performance patterns, identifies missing tools or workflows, and proactively generates new specialized agents and skills to handle complex or repetitive tasks. By maintaining an evolution log and a dedicated 'lessons learned' database, it ensures that every task failure leads to a permanent architectural improvement, transforming Claude from a static tool into a truly adaptive development partner that builds its own infrastructure as it works.

主要功能

01Import system for external agents and skills via GitHub integration

02Automated generation of specialized agents using Haiku, Sonnet, or Opus models

03Proactive Evolution Protocol to detect and suggest capability gaps

04Lessons-learned database for failure analysis and systemic prevention

05Dynamic skill creation for recurring task patterns and automation

0621 GitHub stars

使用场景

01Automating recurring manual steps into permanent, reusable skills

02Systematizing failure recovery by logging root causes and preventing repeat errors

03Scaling complex workflows by creating dedicated sub-agents for specific project architectures

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add kks0488/vibe-claude v-evolve

For use in Claude.ai and ChatGPT