How to Install the AI Engineering from Scratch Skills

You can install the AI Engineering from Scratch skills by running the scripts/install_skills.py script, which automatically validates and copies skill bundles from the repository to ~/.ai-engineering-skills or a custom path defined by the SKILL_ROOT environment variable.

The rohitg00/ai-engineering-from-scratch repository provides a modular collection of reusable skill bundles designed for AI engineering workflows. To leverage these skills in your local environment, you must execute the automated installer that ships with the codebase. This guide explains how to install the AI Engineering from Scratch skills, configure custom installation paths, and verify your setup.

Prerequisites

Before running the installer, ensure you have Python 3 installed. The project relies primarily on the Python standard library, with only minimal third-party dependencies such as numpy listed in requirements.txt.

Step-by-Step Installation

Clone the Repository

First, download the source code from GitHub:

git clone https://github.com/rohitg00/ai-engineering-from-scratch.git
cd ai-engineering-from-scratch

Configure the Python Environment

Create a virtual environment and install the required packages to ensure isolated dependencies:

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Run the Installer Script

Execute scripts/install_skills.py to discover, validate, and install all skill bundles:

python3 scripts/install_skills.py

By default, this command copies validated skill directories (such as skills/start-learning/) from the repository into your home directory at ~/.ai-engineering-skills.

Set the SKILL_ROOT Environment Variable (Optional)

To install skills to a custom location rather than the default home directory path, export the SKILL_ROOT variable before running the installer:

export SKILL_ROOT=/path/to/custom/skills/dir
python3 scripts/install_skills.py

The installer and subsequent helper scripts automatically read this variable to locate your skill bundles.

Verify the Installation

Confirm that the skills are properly installed by listing the contents of the skill root directory:

ls $SKILL_ROOT

You should see directories like start-learning, learn, and learn-mcp, each containing a SKILL.md manifest file.

How the Skill Installer Works

The scripts/install_skills.py script performs four distinct operations to ensure a clean, valid installation:

• Discovery: The script walks the skills/ directory tree to locate every SKILL.md manifest file.
• Validation: It checks each manifest for required metadata fields, including title, description, and usage sections.
• Copying: It creates a clean target directory at $SKILL_ROOT (defaulting to ~/.ai-engineering-skills) and copies the complete skill bundles—including SKILL.md and any supporting files—into this location.
• Indexing: The script generates an index.json file in the root of the installation directory, creating a machine-readable registry so that higher-level scripts can locate specific skills by name without scanning the filesystem.

Because the installer is pure Python and only manipulates files within the repository boundary, it respects the repository's strict no-secret and no-external-execution security policies.

Programmatically Accessing Installed Skills

After installation, you can load skill manifests directly from Python using the SKILL_ROOT environment variable:

import json
import os
import pathlib

# Resolve the skill root using the environment variable or default path

skill_root = pathlib.Path(
    os.getenv("SKILL_ROOT", "~/.ai-engineering-skills")
).expanduser()

# Load the "start-learning" skill manifest

manifest_path = skill_root / "start-learning" / "SKILL.md"
with open(manifest_path, "r", encoding="utf-8") as f:
    content = f.read()

print(content[:500])  # Preview the first 500 characters

This approach allows you to integrate skill documentation and metadata into custom CLI tools or Jupyter notebooks.

Summary

• Run python3 scripts/install_skills.py to install all skills from the repository.
• The default installation path is ~/.ai-engineering-skills, configurable via the SKILL_ROOT environment variable.
• The installer validates each SKILL.md manifest in the skills/ directory before copying bundles to the target location.
• An index.json file is generated to enable fast skill lookup by name.
• Dependencies are minimal; only standard library packages and numpy (specified in requirements.txt) are required.

Frequently Asked Questions

Where are the AI Engineering from Scratch skills installed by default?

By default, the installer places skills in ~/.ai-engineering-skills (a hidden directory in your home folder). This path contains subdirectories for each skill, such as start-learning/, each housing a SKILL.md manifest and related files.

Can I change the installation directory for the skills?

Yes. Set the SKILL_ROOT environment variable to your desired path before running scripts/install_skills.py. The installer will use this variable as the target directory, and subsequent scripts will reference it to locate your skills.

What Python dependencies are required to run the installer?

The installer requires only the Python standard library plus packages listed in requirements.txt, which includes numpy and other common data science utilities. No external API keys or secret management tools are needed, adhering to the repository's security guidelines.

How does the installer validate skill bundles before installation?

The script checks each SKILL.md file found in the skills/ directory for mandatory fields including title, description, and usage. If a manifest is missing required metadata, the installer skips that bundle and logs a validation error, ensuring only complete, documented skills are copied to your local environment.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}