super-kit

Super-Kit

Super-Kit is a modular repository containing instructions, workflows, skills, and specializations for AI coding agents. By structuring AI agent contexts into individual files, it allows large language model agents to load specific knowledge on the fly instead of relying on massive and bloated static system prompts.

While the primary purpose is to compund the knowledge of engineering, I will add more domains as I see fit.

🔗 GitHub Repository: dgkngk/super-kit

🔗 NPM Package: superkit-mcp-server

Install as Claude Code Plugin

1. Open Claude Code → /plugin → Discover → Add source
2. Enter: github:dgkngk/super-kit
3. Install: /plugin install super-kit@dgkngk

This configures the MCP server automatically and makes all skills, agents, and commands available.

Requirements: Node.js 18+ (for npx superkit-mcp-server@latest)

Directory Structure

• agents/: Contains instructions and guidelines for specialized AI roles (e.g., data-engineer).
• skills/: Contains technology-specific or meta skills (patterns, best practices) the agent can load dynamically (e.g., react-best-practices).
• workflows/: Contains step-by-step interactive slash-commands to guide the AI workflow (e.g., /plan, /explore).
• src/: The Model Context Protocol (MCP) server that exposes all these assets directly to compatible AI platforms.

Providing the Kit to Your AI

The easiest way to power up your agent with this toolkit is to use the included MCP server.

You can launch it directly using npx:

npx -y superkit-mcp-server

Available Tools

• list_superkit_assets: Lists all available agents, skills (tech/meta), and workflows in the Super-Kit repository.
• load_superkit_agent: Loads the system instructions and guidelines for a specific specialist agent (e.g., data-engineer).
• load_superkit_skill: Loads the skill instructions (SKILL.md) for a specific category and skill (e.g., category: tech, name: react-best-practices).
• load_superkit_workflow: Loads the instructions for a specific slash-command workflow (e.g., work, explore).
• call_tool_checklist: Executes the native TypeScript validation suite (security, web accessibility, react performance, testing, API structure) on a target project location via the MCP environment instead of generic bash loops.

Available Prompts

The Super-Kit MCP server exposes all workflows inside the workflows/ directory dynamically as MCP Prompts. AI agents can invoke GetPromptRequestSchema to instantly load complex workflows (e.g., plan-compound, review-compound) along with their security and validation tooling straight into their context.

Manual Installation and Configuration

If you cloned this repository locally, you can build and use the MCP server directly:

cd super-kit
npm install
npm run build

Configuring in AI Platforms

Cline / Roo (VS Code)

Add the following to your cline_mcp_settings.json:

{
  "mcpServers": {
    "superkit": {
      "command": "node",
      "args": ["/absolute/path/to/super-kit/build/index.js"]
    }
  }
}

Cursor / Windsurf

Go to Settings -> Features -> MCP Servers. Add a new stdio server:

• Name: superkit
• Command: node /absolute/path/to/super-kit/build/index.js

How it works

The built-in MCP server reads directly from the super-kit directory, resolving paths safely to prevent traversal attacks. It automatically parses and returns the Markdown contents of the structural elements so your AI agent always has the right context in mind.