Initialize AI coding environments for your projects.
📖 中文文档
Setting up AI coding agents from scratch means manually choosing MCP servers, skills, and configuration files for each project. agent-init automates this: detect your stack, pick the right tools, and generate ready-to-use config -- so you can skip the boilerplate and start coding.
agent-init is a skill that detects your project and configures an optimal environment for AI coding agents. When invoked, it:
- Detects your project type (language, framework, build tools)
- Recommends MCP servers and skills matched to your stack
- Generates configuration files (
.mcp.json,AGENTS.md) - Optionally suggests heavy tools (CodeGraph, RTK, Superpowers) for you to approve
Install the init-harness skill with npx skills:
npx skills add micln/agent-initAfter install, run it in Claude:
/init-harness
After install, reference init-harness in your OpenCode config and invoke it.
| Type | Profile | Key MCP Servers |
|---|---|---|
| Generic | base.md |
context7, github |
| Next.js | nextjs.md |
context7, github, next-devtools, playwright |
| React | react.md |
context7, github, playwright |
| Godot | godot.md |
context7, github, godot-mcp |
| Go | go.md |
context7, github |
These are high-impact developer tools that require user confirmation before installation:
| Tool | Purpose | Install |
|---|---|---|
| CodeGraph | Code knowledge graph (59K stars) | npx @colbymchenry/codegraph |
| RTK | Token compression (70K stars) | cargo install --git https://github.com/rtk-ai/rtk |
| Superpowers | Development methodology (253K stars) | See docs |
MIT