install-labs

# Install Labs v2.0.0

**Turn your AI agents and automations into installable plugins and packages.**

You built an agent. Now make it so other people can install and use it.

[![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
[![Claude Code Plugin](https://img.shields.io/badge/Claude_Code-Plugin-blueviolet)](https://docs.anthropic.com/en/docs/claude-code)
[![Version](https://img.shields.io/badge/version-2.0.0-green.svg)]()

---

## Table of Contents

- [What This Plugin Does](#what-this-plugin-does)
- [The Problem](#the-problem)
- [The Journey](#the-journey)
- [Commands (10)](#commands-10)
- [Skills (12)](#skills-12)
- [Frameworks Covered (17)](#frameworks-covered-17)
- [Distribution Targets (10)](#distribution-targets-10)
- [Academic Foundations](#academic-foundations)
- [Quick Start](#quick-start)
- [Installation](#installation)
- [Architecture](#architecture)
- [References](#references)
- [License](#license)

---

## What This Plugin Does

Install Labs is a Claude Code plugin that guides you through packaging any AI agent or automation -- built with any framework -- into an installable product. It covers every major distribution target: MCP servers, Claude Code plugins, Custom GPTs, Docker containers, PyPI packages, npm packages, and cloud deployments.

**One mission:** "I built an AI agent. Help me turn it into something other people can install and use."

---

## The Problem

AI agents are harder to package than traditional software. Traditional software has a single runtime, a known dependency tree, and a build artifact. AI agents break every assumption:

| Challenge | Traditional Software | AI Agents |
|---|---|---|
| **Secrets** | Maybe one DB connection string | 3-8 API keys across providers |
| **Model weights** | No large binary blobs | 500MB-70GB model files |
| **Runtime environment** | One language runtime | Python + Node.js + system libs + CUDA/Metal |
| **Config sprawl** | One config file | `.env`, `config.yaml`, `plugin.json`, `mcp.json`, framework configs |
| **Framework churn** | Stable APIs | LangChain, CrewAI, AutoGen -- breaking changes monthly |
| **Hardware variance** | CPU is CPU | GPU type, VRAM, quantization all affect behavior |
| **State** | Database handles it | Vector stores, conversation memory, tool caches |

The result: an agent that works on the author's machine silently depends on 15 things the author forgot they installed.

Research in cognitive load theory (Sweller, 1988) shows that every additional install step increases the probability of user abandonment. The Fogg Behavior Model (Fogg, 2009) predicts that install completion requires the intersection of motivation (wanting the agent), ability (low friction install), and trigger (clear call to action). When any of these three factors is insufficient, the user drops off. Install Labs applies these principles systematically to every distribution target.

---

## The Journey

```
+---------------------------------------------------------------------------+
|                    INSTALL LABS: AGENT -> INSTALL                          |
|                    10 Commands . 3 Phases . 12 Skills                     |
|                                                                           |
|  -- PHASE 0: ASSESS ---------------------------------------------------- |
|     Figure out what you have and where it should go                       |
|                                                                           |
|     /agent-guide  (0.1)  Orientation & routing                           |
|     /agent-audit  (0.2)  Packaging readiness audit                       |
|                                                                           |
|  -- PHASE 1: PACKAGE --------------------------------------------------- |
|     Choose a distribution target and generate packaging files             |
|                                                                           |
|     /pick-target    (1.1)  Choose: MCP, plugin, GPT, Docker, etc.       |
|     /pick-framework (1.2)  Framework-specific packaging guide            |
|     /gen-package    (1.3)  Generate all packaging files                  |
|                                                                           |
|  -- PHASE 2: SHIP ------------------------------------------------------ |
|     Validate, secure, and release                                         |
|                                                                           |
|     /gen-install  (2.1)  Generate install script/instructions            |
|     /pre-ship     (2.2)  Agent-specific pre-ship checklist               |
|     /secure       (2.3)  Agent security audit                            |
|                                                                           |
|  -- STANDALONE ---------------------------------------------------------- |
|     /agent-roast         Critique any agent's install experience         |
|     /glossary            Agent packaging terminology                     |
|                                                                           |
|  -- FLOW ---------------------------------------------------------------- |
|     ASSESS ---------> PACKAGE ---------> SHIP                            |
|     What do you       Turn it into       Validate &                      |
|     have?             an installable     release                         |
|     (2 commands)      (3 commands)       (3 commands) + 2 standalone     |
+---------------------------------------------------------------------------+
```

---

## Commands (10)

### Phase 0: ASSESS
| Command | Step | What It Does |
|---------|------|-------------|
| `/agent-guide` | 0.1 | Orientation -- identify your agent, get routed to the right path |
| `/agent-audit` | 0.2 | Audit your agent's packaging readiness across 6 dimensions (30-point scoring) |

### Phase 1: PACKAGE
| Command | Step | What It Does |
|---------|------|-------------|
| `/pick-target` | 1.1 | Interactive decision tree for choosing distribution target: MCP, plugin, GPT, Docker, PyPI, npm, cloud |
| `/pick-framework` | 1.2 | Framework-specific packaging guide for your stack (17 frameworks x 8 targets) |
| `/gen-package` | 1.3 | Generate all packaging files -- pyproject.toml, package.json, Dockerfile, plugin.json, etc. |

### Phase 2: SHIP
| Command | Step | What It Does |
|---------|------|-------------|
| `/gen-install` | 2.1 | Generate production install.sh/install.ps1 scripts and README install sections |
| `/pre-ship` | 2.2 | Run 39-item pre-ship quality gate across 7 sections with pass/fail verdicts |
| `/secure` | 2.3 | 10-point agent security audit (API keys, supply chain, permissions, model integrity) |

### Standalone
| Command | What It Does |
|---------|-------------|
| `/agent-roast` | Critique any agent's install experience across 8 dimensions with letter grade (A-F) |
| `/glossary` | Plain-language definitions of 40+ agent packaging terms with code examples |

---

## Skills (12)

Skills are auto-invoked when the conversation matches their domain. Each skill contains production-ready code examples, decision frameworks, and citations to official documentation.

| Skill | Lines | Coverage |
|-------|-------|----------|
| **agent-packaging-foundations** | 229 | Core packaging principles, dependency taxonomy, install funnel, 8 anti-patterns |
| **mcp-server-packaging** | 640 | MCP specification, 6 distribution methods, config injection, registries, testing |
| **claude-code-plugin-packaging** | 590 | Plugin architecture, skills, commands, CLAUDE.md, distribution methods |
| **custom-gpt-packaging** | 428 | GPT Builder, Actions (OpenAPI 3.1), knowledge files, GPT Store publishing |
| **docker-agent-packaging** | 804 | Multi-stage builds, GPU support, one-click deploy, 5 cloud platforms |
| **pypi-agent-packaging** | 837 | pyproject.toml, pip/uvx, lazy imports, Trusted Publishers, CI/CD |
| **npm-agent-packaging** | 614 | package.json, npx z

[truncated…]