Building a Secure Infinito.Nexus AI Development Environment on Linux

Development AI
1

AI coding agents like Claude Code, OpenAI Codex, and Gemini CLI are transforming software development. They can:

  • edit files autonomously
  • execute shell commands
  • refactor projects
  • install dependencies
  • generate infrastructure code
  • manage workflows across repositories

That power also creates a major security risk.

If you run these agents directly on your host operating system, a bad prompt, hallucinated command, broken plugin, or compromised dependency can quickly damage your system.

This guide explains how to build a secure AI development environment for Infinito.Nexus using:

  • a dedicated Linux VM
  • Code OSS
  • Claude Code
  • OpenAI Codex
  • Gemini integrations
  • isolated AI tooling
  • Git-based safety workflows

The goal is simple:

Never run autonomous AI agents directly on your production workstation or host operating system.

Why You Should Always Use a VM

AI coding agents are not simple chatbots anymore.

Modern agents can:

  • recursively edit files
  • execute bash commands
  • install packages
  • modify Docker containers
  • rewrite configurations
  • delete directories
  • access secrets accidentally
  • perform autonomous multi-step workflows

Even with safety systems, mistakes happen.

Infinito.Nexus itself includes explicit AI-agent governance documentation through:

AGENTS.md
CLAUDE.md
GEMINI.md

These files define:

  • sandbox behavior
  • permission boundaries
  • allowed commands
  • restricted operations
  • workflow conventions

(github.com) (GitHub)

The safest setup is therefore:

Host Operating System
└── Virtual Machine
    ├── Linux
    ├── Code OSS
    ├── AI Agents
    ├── Docker
    └── Infinito.Nexus Development Environment

If an agent breaks the VM:

  • your host remains safe
  • secrets stay isolated
  • snapshots allow instant rollback
  • experiments remain disposable

Recommended Linux Distributions

Any modern Linux distribution works well.

Recommended choices:

Distribution Why It Works Well
Fedora Workstation Excellent developer experience
Ubuntu LTS Large ecosystem and tutorials
Debian Stable and predictable
Arch Linux Cutting-edge tooling
CentOS Stream Enterprise-oriented workflows

For most users:

  • Fedora Workstation
  • Ubuntu 24.04 LTS

are the best choices.

Recommended Virtualization Platforms

Use one of the following:

Platform Notes
KVM/QEMU Best native Linux performance
VirtualBox Easy cross-platform usage
VMware Workstation Stable enterprise solution
GNOME Boxes Beginner friendly
Proxmox Excellent for dedicated lab environments

Recommended VM resources:

Resource Recommendation
RAM 16 GB
CPU 4+ cores
Disk 100 GB SSD
Network NAT

Install Base Development Dependencies

Inside the VM install the required tooling.

Fedora

sudo dnf install -y \
    git \
    make \
    curl \
    wget \
    nodejs \
    npm \
    python3 \
    python3-pip \
    docker \
    docker-compose \
    ansible

Ubuntu / Debian

sudo apt update

sudo apt install -y \
    git \
    make \
    curl \
    wget \
    nodejs \
    npm \
    python3 \
    python3-pip \
    docker.io \
    docker-compose \
    ansible

Enable Docker:

sudo systemctl enable --now docker
sudo usermod -aG docker $USER

Relogin afterwards.

Install Code OSS

Code OSS is the open-source foundation of Visual Studio Code.

Fedora

sudo dnf install code

Ubuntu / Debian

sudo snap install code --classic

Alternative distributions:

  • VSCodium
  • official VSCode packages

Downloads:

Clone Infinito.Nexus Core

Clone the repository:

git clone https://github.com/infinito-nexus/core.git

Enter the repository:

cd core

Repository:

https://github.com/infinito-nexus/core

Setup the Environment

The project uses make for environment automation.

Initial setup:

make

Useful commands:

make help
make install
make setup
make lint
make test

The Infinito.Nexus ecosystem heavily relies on:

  • Docker
  • Linux
  • Ansible
  • modular automation
  • ActivityPub
  • LDAP-based IAM
  • Keycloak SSO

(github.com) (GitHub)

Understanding AI Agent Configuration Files

Modern AI coding tools automatically load repository instructions from markdown configuration files.

Important standards:

File Used By
AGENTS.md Codex, Cursor, Claude, others
CLAUDE.md Claude Code
GEMINI.md Gemini CLI
.github/copilot-instructions.md GitHub Copilot
.cursorrules Cursor

The industry is increasingly converging around AGENTS.md as the universal standard. (Agents)

These files define:

  • project structure
  • coding conventions
  • safety rules
  • testing workflows
  • forbidden actions
  • command usage

Before working with AI agents:

cat AGENTS.md

You should always understand:

  • what the agent may execute
  • what it may modify
  • where secrets must never be exposed
  • how sandboxing behaves

Installing Claude Code

Install Claude Code globally:

npm install -g @anthropic-ai/claude-code

Launch:

claude

Claude Code automatically reads:

CLAUDE.md
AGENTS.md

depending on repository configuration. (DeployHQ)

Installing OpenAI Codex

Install Codex:

npm install -g @openai/codex

Or:

curl -fsSL https://chatgpt.com/codex/install.sh | sh

Run:

codex

Codex primarily reads:

AGENTS.md

and supports hierarchical agent configuration. (DeployHQ)

Installing Gemini CLI

Install Gemini tooling:

npm install -g @google/gemini-cli

Run:

gemini

Gemini automatically reads:

GEMINI.md

within the repository. (DeployHQ)

Installing AI Agent Extensions in Code OSS

One of the most important steps is properly integrating AI agents into Code OSS.

Open the Extensions view:

Ctrl + Shift + X

or:

View → Extensions

Install Claude Extensions

Search for:

Claude Code

Recommended extensions:

Extension Purpose
Claude Code Claude agent integration
Anthropic Tools Claude workflows and utilities

After installation:

  1. Open terminal
  2. Authenticate:
claude login
  1. Restart Code OSS

Claude extensions will automatically detect:

CLAUDE.md
AGENTS.md

inside the repository.

Install OpenAI Codex Extensions

Search for:

Codex
OpenAI

Recommended extensions:

Extension Purpose
OpenAI Codex Codex integration
OpenAI Tools Chat and workflow tooling

After installation:

  1. Open terminal
  2. Authenticate:
codex login

or configure:

export OPENAI_API_KEY=...

Codex automatically reads:

AGENTS.md

for repository guidance.

Install Gemini Extensions

Search for:

Gemini
Google

Recommended extensions:

Extension Purpose
Gemini Code Assist AI coding assistance
Gemini Chat Gemini interaction panel

After installation:

  1. Sign in with Google
  2. Restart Code OSS
  3. Open the repository folder

Gemini automatically loads:

GEMINI.md

from the repository.

Recommended Additional Extensions

Recommended developer extensions:

Extension Purpose
GitLens Git visualization
Docker Container tooling
YAML YAML support
Python Python tooling
Ansible Ansible support
Markdown All In One Better markdown editing
Error Lens Inline diagnostics

Recommended AI Agent Workspace Layout

A good secure AI workspace looks like this:

~/workspace/
├── core/
├── projects/
├── sandbox/
├── experiments/
└── backups/

Never give agents unrestricted access to:

  • your home directory
  • SSH keys
  • browser profiles
  • host filesystem
  • production secrets

Secure Workflow Recommendations

1. Commit Frequently

AI agents can create massive changes very quickly.

Commit constantly:

git add .
git commit -m "checkpoint"

2. Snapshot the VM Frequently

Always create snapshots before:

  • dependency upgrades
  • autonomous refactors
  • infrastructure changes
  • large AI-generated migrations

3. Never Expose Secrets in Prompts

Never paste:

  • passwords
  • API keys
  • production tokens
  • SSH private keys

into AI chats or prompts.

4. Keep the Environment Disposable

Your environment should always be rebuildable.

A healthy AI workflow assumes:

destroy VM → recreate VM → continue working

5. Keep AI Agents Contained

Never allow unrestricted access to:

  • host operating system
  • mounted personal folders
  • password stores
  • production infrastructure

Treat AI agents like semi-autonomous infrastructure processes.

Useful Infinito.Nexus Resources

Repository

https://github.com/infinito-nexus/core

AI Cheatsheet

https://s.infinito.nexus/cheatsheet

Agent Documentation

AGENTS.md
CLAUDE.md
GEMINI.md

Example Daily Workflow

Start VM:

virsh start infinito-dev

Open repository:

cd ~/workspace/core

Start Code OSS:

code .

Start Claude:

claude

Or Codex:

codex

Or Gemini:

gemini

Review all changes carefully:

git diff

Commit validated work:

git commit -am "validated AI agent changes"

Final Thoughts

AI-assisted development is incredibly powerful, but secure isolation is now mandatory.

The safest setup combines:

  • Linux VMs
  • isolated development environments
  • Code OSS
  • Git workflows
  • sandboxed AI agents
  • disposable infrastructure

Infinito.Nexus already embraces structured AI governance using:

  • AGENTS.md
  • CLAUDE.md
  • GEMINI.md

making it an excellent foundation for secure AI-assisted infrastructure engineering. (GitHub)