Auto-Installer 🤖

Autonomous AI agent that installs ANYTHING on any system with zero manual intervention.

Handles system packages, npm/pip/cargo packages, API configurations, environment setup, VM deployments, and more.

Quick Start

git clone https://github.com/anombyte93/auto-installer.git
cd auto-installer

Then use with any AI tool below ⬇️

🤖 AI Tool Compatibility

Tool	Type	Installation	Status
Claude Code	CLI Skill	cp -r . ~/.claude/skills/auto-installer/	✅ Native
Open Codex	CLI Agent	See below	✅ Full
Gemini CLI	CLI Agent	See below	✅ Full
Grok CLI	CLI Agent	See below	✅ Full
OpenAI Codex	CLI Agent	See below	✅ Full
Continue.dev	IDE Extension	VSCode/JetBrains	✅ Full
Aider	CLI Agent	See below	✅ Full
ChatGPT CLI	CLI	N/A	✅ Full
Any AI with shell access	Custom	Paste SKILL.md	✅ Universal

📦 Installation by Tool

Claude Code (Skill System)

# Install as a skill
cp -r . ~/.claude/skills/auto-installer/

# Auto-activates when you say:
# "install docker"
# "configure gemini API in open-codex"
# "install pytest on my ubuntu VM"

CLI Tools (Open Codex, Gemini CLI, Grok CLI, Codex, Aider)

All CLI tools follow the same pattern:

cd auto-installer

# Start your AI CLI tool
open-codex          # or
gemini-cli          # or
grok-cli            # or
codex               # or
aider

# Tell the AI:
"Read SKILL.md and install [WHAT YOU WANT]"

Tool-Specific Context Flags (if supported):

# Gemini CLI
gemini-cli --context SKILL.md

# Some tools auto-read files in directory
open-codex  # Automatically detects SKILL.md

Continue.dev (IDE Extension)

1. Install Continue extension in VSCode/JetBrains
2. Open the auto-installer directory in your IDE
3. Use Continue chat: @SKILL.md install docker

Universal Method (Any AI)

Paste this prompt to any AI with shell access:

You are an autonomous software installer. Read the file SKILL.md in the current directory to understand your capabilities and workflow.

CRITICAL RULES:
1. Execute ALL commands yourself - NEVER give me manual instructions
2. Only ask for: passwords, disambiguation, or risky operation confirmations
3. Report "READY TO USE" only when fully installed and verified

Now install: [WHAT YOU WANT TO INSTALL]

🎯 What It Does

✅ Auto-detects installation type (system pkg, npm, API config, etc.) ✅ System packages - docker, postgresql, redis, nginx, etc. ✅ Language packages - npm, pip, cargo, gem, composer, go ✅ API configurations - Set API keys, create config files ✅ Environment setup - PATH, .bashrc, .env files ✅ VMs & remote servers - Auto-configures SSH, jump hosts ✅ GPU drivers - NVIDIA drivers + nvidia-docker with reboot ✅ Safety checkpoints - VM snapshots before changes ✅ Error recovery - Automatic retry with fixes ✅ Verification - Functional tests, not just binary checks

💡 Examples

System Packages

install docker
install node 20 and postgres 15
install rust

NPM/Language Packages

install npm package open-codex
install pytest and numpy
install ruby gem rails

API & Configuration

configure gemini API in open-codex
set up my AWS credentials
create .env file with DATABASE_URL

VM & Remote

install pytest on my ubuntu VM
install docker on staging and production VMs
install rust on gpu-server.company.com via bastion.company.com

GPU & ML

install docker with GPU support and pytorch on my GPU server

📚 Supported Installation Types

System Packages

• Package Managers: apt, brew, dnf, pacman, choco
• Examples: docker, postgresql, redis, nginx, git, curl, vim

Language Package Managers

• Python: pip, pip3, pipx, poetry
• Node.js: npm, yarn, pnpm, bun
• Ruby: gem, bundler
• Rust: cargo, rustup
• Go: go install, go get
• PHP: composer
• Perl: cpan
• Java: maven, gradle

Development Tools

• Version Managers: nvm, pyenv, rbenv
• CLI Tools: gh, kubectl, terraform, ansible, aws-cli
• IDEs: VSCode extensions (code --install-extension)
• Desktop Apps: snap, flatpak, AppImage, Homebrew Cask

Configuration & Setup

• API Keys: Environment variables, config files
• Config Files: JSON, TOML, YAML creation/editing
• Environment: PATH, .bashrc, .zshrc, .env files
• SSH Keys: Generation and configuration

Deployment Targets

• Local system - Direct installation
• VMs - VirtualBox, VMware, Parallels (auto-detects)
• Remote servers - Via SSH
• Jump hosts - Multi-hop SSH configuration
• Containers - Docker, Kubernetes pods

Specialized

• GPU Drivers: NVIDIA drivers + nvidia-docker
• ML/Data Science: PyTorch, TensorFlow, CUDA
• Language Runtimes: python, node, ruby, go, rust, java, php

🔧 How It Works

1. Parse - AI detects package type and target system
2. Configure - Sets up SSH/VMs if needed
3. Checkpoint - Creates VM snapshot or system backup
4. Install - Uses appropriate package manager
5. Verify - Runs functional tests (not just checks if binary exists)
6. Report - Confirms "READY TO USE" with versions

🛡️ Safety Features

• VM snapshots before any changes
• System state backups for remote servers
• Checkpoint at every critical step
• Automatic rollback on failures
• Never leaves systems in broken state
• Logs all actions for debugging

⚙️ Requirements

• Any OS: Linux, macOS, Windows/WSL
• For VMs: VirtualBox, VMware, or Parallels
• For Remote: SSH access (auto-configured if needed)
• For GPU: NVIDIA GPU (drivers auto-installed)

🔍 Troubleshooting

AI asks me to run commands manually:

• Emphasize: "Execute everything yourself. I cannot run commands."
• Add to prompt: "CRITICAL: Run all commands yourself via shell access"

AI doesn't read SKILL.md:

• Use: "Read the file SKILL.md then install XYZ"
• Or paste SKILL.md content directly into chat

Skill doesn't activate in Claude Code:

• Check it's in ~/.claude/skills/auto-installer/SKILL.md
• Verify YAML frontmatter is valid
• Say explicitly: "use auto-installer skill to install XYZ"

Installation logs:

• Check: /var/log/apt/, ~/.npm/_logs/, ~/.pip/
• All bash commands are logged by Claude/AI tools
• Use history to review executed commands

📁 File Structure

auto-installer/
├── SKILL.md          # Complete installation logic (AI reads this)
├── README.md         # This file (for humans)
├── EXAMPLES.md       # Real-world usage examples
├── CONTRIBUTING.md   # How to contribute
├── CHANGELOG.md      # Version history
└── LICENSE           # MIT License

🚀 Advanced Usage

Multi-Package Installations

install docker, node 20, postgresql, redis, and python testing tools

The AI will:

• Clarify ambiguities ("python testing tools" → pytest only? full suite?)
• Install everything in optimal dependency order
• Configure and start all services
• Verify each package works

Production Deployments

install docker on prod-server.example.com

The AI will:

• Create system backups first
• Install packages with minimal downtime
• Start and verify services
• Keep backups until success confirmed

GPU Servers

install docker with nvidia GPU support and pytorch with CUDA on gpu-server

The AI will:

• Detect GPU hardware
• Install NVIDIA drivers
• Reboot server and poll until back online
• Install nvidia-docker
• Install PyTorch with matching CUDA version
• Verify GPU access with test container

🤝 Contributing

See CONTRIBUTING.md for guidelines on:

• Reporting bugs
• Suggesting features
• Submitting improvements
• Testing new scenarios

📜 License

MIT License - Use freely with any AI tool

💭 Philosophy

Core Principles:

• Correctness > Speed - Do it right, not fast
• Safety > Convenience - Checkpoints before changes
• Autonomy > User interaction - ZERO manual commands
• Transparency > Abstraction - Clear logs and reports
• Robustness > Simplicity - Handle edge cases

Design Goals:

• ZERO manual commands for the user
• Only ask for: passwords, disambiguation, risky confirmations
• Always verify installations actually work
• Create safety checkpoints before changes
• Recover from errors automatically
• Never leave systems in broken state

---

Made with ❤️ for the AI coding community

Supports: Claude Code • Open Codex • Gemini CLI • Grok CLI • Continue.dev • Aider • and more