Production-ready TypeScript backend development with Bun runtime. Includes specialized agents for backend development, API design, and DevOps. Features comprehensive best practices, tools integration (Biome, Prisma, Hono, Docker), testing workflows, and AWS ECS deployment guidance (2025).

Installation

npx claude-plugins install @MadAppGang/mag-claude-plugins/bun

Contents

Folders: agents, commands, mcp-servers, skills

Files: ARCHITECTURE.md, QUICK_REFERENCE_SPEC.md, README.md, TIERED_PRICING_SPEC.md, plugin.json, recommended-models.md

Documentation

This directory contains resources shared across all Claude Code plugins.

Purpose

Centralized management of common resources to ensure consistency and reduce duplication.

Architecture

shared/                                    ← SOURCE OF TRUTH (edit here)
├── recommended-models.md                  ← Model recommendations
├── QUICK_REFERENCE_SPEC.md                ← Quick reference spec
├── README.md                              ← This file
└── skills/                                ← Shared skills
    └── claudish-usage/                    ← Claudish usage skill
        └── SKILL.md                       ← AUTO-DISTRIBUTED

scripts/
└── sync-shared.ts                         ← Distribution script (recursive)

plugins/
├── frontend/
│   ├── recommended-models.md              ← AUTO-COPIED
│   ├── QUICK_REFERENCE_SPEC.md            ← AUTO-COPIED
│   ├── README.md                          ← AUTO-COPIED
│   └── skills/                            ← AUTO-CREATED
│       └── claudish-usage/                ← AUTO-CREATED
│           └── SKILL.md                   ← AUTO-COPIED
├── bun/
│   ├── recommended-models.md              ← AUTO-COPIED
│   ├── QUICK_REFERENCE_SPEC.md            ← AUTO-COPIED
│   ├── README.md                          ← AUTO-COPIED
│   └── skills/                            ← AUTO-CREATED
│       └── claudish-usage/                ← AUTO-CREATED
│           └── SKILL.md                   ← AUTO-COPIED
└── code-analysis/
    ├── recommended-models.md              ← AUTO-COPIED
    ├── QUICK_REFERENCE_SPEC.md            ← AUTO-COPIED
    ├── README.md                          ← AUTO-COPIED
    └── skills/                            ← AUTO-CREATED
        └── claudish-usage/                ← AUTO-CREATED
            └── SKILL.md                   ← AUTO-COPIED

How It Works

1. Edit Source Files

All shared resources are stored in shared/ directory. This is the ONLY place you should edit these files.

Example:

# Edit the source
vim shared/recommended-models.md

2. Sync to Plugins

Run the sync script to copy resources to all plugins:

# Sync all shared resources
bun run sync-shared

# Or use the short alias
bun run sync

3. Automatic Distribution

The sync script copies each file from shared/ to all plugin directories:

• shared/recommended-models.md → plugins/frontend/recommended-models.md
• shared/recommended-models.md → plugins/bun/recommended-models.md
• shared/recommended-models.md → plugins/code-analysis/recommended-models.md

4. Plugin Usage

Commands and agents read the synced files using plugin-relative paths:

For flat files:

Read file: ${CLAUDE_PLUGIN_ROOT}/recommended-models.md

For skills:

Read file: ${CLAUDE_PLUGIN_ROOT}/skills/claudish-usage/SKILL.md

This ensures each plugin always has access to the latest recommendations and shared skills.

Shared Skills

What are Shared Skills?

Shared skills are reusable knowledge modules that are distributed across all plugins. They provide:

• ✅ Consistent guidance across plugins
• ✅ Centralized knowledge management
• ✅ Automatic updates when skills are improved
• ✅ Reduced duplication

Current Shared Skills

1. claudish-usage (skills/claudish-usage/SKILL.md)
   • How to use Claudish CLI with OpenRouter models
   • File-based sub-agent patterns
   • Model selection and cost tracking
   • Best practices for AI agents

Adding New Shared Skills

1. Create skill directory: shared/skills/skill-name/
2. Add SKILL.md file with YAML frontmatter:

   ---
   name: skill-name
   description: Clear description of what the skill does and when to use it
   ---
   

3. Write skill content (instructions, examples, best practices)
4. Run bun run sync-shared
5. Skill is automatically distributed to all plugins and discovered by Claude

Using Shared Skills in Commands/Agents

Skills are automatically discovered - Claude loads them based on the task at hand.

Skills must have YAML frontmatter with name and description:

---
name: claudish-usage
description: Guide for using Claudish CLI to run Claude Code with OpenRouter models. Use when working with external AI models, multi-model workflows, or cost optimization.
---

How it works:

1. Startup: Skill metadata (name + description) loaded into system prompt (~100 tokens/skill)
2. Task matching: When user request matches description, Claude automatically reads SKILL.md via bash
3. Progressive loading: Claude only loads relevant skills, avoiding context pollution

You do NOT need to:

• ❌ Manually read skill files with Read tool
• ❌ Reference skills in agent/command prompts
• ❌ Load skills into context explicitly

Claude automatically uses skills when:

• ✅ Task matches the skill’s description
• ✅ User mentions keywords from description (e.g., “Claudish”, “OpenRouter”, “Grok”)
• ✅ Workflow requires skill’s cap

…(truncated)

Included Skills

This plugin includes 2 skill definitions:

best-practices

Use when implementing production-ready TypeScript backend applications with Bun runtime, including project structure, performance optimization, error handling, testing, and deployment patterns.

project-root/
├── src/
│   ├── server.ts              # Entry point (starts server)
│   ├── app.ts                 # Hono app initialization & middleware
│   ├── config.ts              # Environment configuration
│   ├── core/ 

...(truncated)

</details>

### claudish-usage

> CRITICAL - Guide for using Claudish CLI ONLY through sub-agents to run Claude Code with OpenRouter models (Grok, GPT-5, Gemini, MiniMax). NEVER run Claudish directly in main context unless user explicitly requests it. Use when user mentions external AI models, Claudish, OpenRouter, or alternative models. Includes mandatory sub-agent delegation patterns, agent selection guide, file-based instructions, and strict rules to prevent context window pollution.

<details>
<summary>View skill definition</summary>

plugin: bun
updated: 2026-01-20

# Claudish Usage Skill

**Version:** 1.1.0
**Purpose:** Guide AI agents on how to use Claudish CLI to run Claude Code with OpenRouter models
**Status:** Production Ready

## ⚠️ CRITICAL RULES - READ FIRST

### 🚫 NEVER Run Claudish from Main Context

**Claudish MUST ONLY be run through sub-agents** unless the user **explicitly** requests direct execution.

**Why:**
- Running Claudish directly pollutes main context with 10K+ tokens (full conversation + reasoning)
- Destroys context window efficiency
- Makes main conversation unmanageable

**When you can run Claudish directly:**
- ✅ User explicitly says "run claudish directly" or "don't use a sub-agent"
- ✅ User is debugging and wants to see full output
- ✅ User specifically requests main context execution

**When you MUST use sub-agent:**
- ✅ User says "use Grok to implement X" (delegate to sub-agent)
- ✅ User says "ask GPT-5 to review X" (delegate to sub-agent)
- ✅ User mentions any model name without "directly" (delegate to sub-agent)
- ✅ Any production task (always delegate)

### 📋 Workflow Decision Tree

## 🤖 Agent Selection Guide

### Step 1: Find the Right Agent

**When user requests Claudish task, follo

...(truncated)

</details>

## Source

[View on GitHub](https://github.com/MadAppGang/claude-code)