# How to Use the start-learning Skill to Begin the AI Engineering from Scratch Curriculum

> Learn to use the start-learning skill to initialize your AI Engineering from Scratch curriculum. Create a personalized LEARNING.md file with your mission and structured learning path.

- Repository: [Rohit Ghumare/ai-engineering-from-scratch](https://github.com/rohitg00/ai-engineering-from-scratch)
- Tags: how-to-guide
- Published: 2026-08-26

---

**The `start-learning` skill initializes your learning journey by creating a personalized [`LEARNING.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/LEARNING.md) file containing your mission statement, skill placement level, and a structured path through 20 curriculum phases.**

As implemented in the `rohitg00/ai-engineering-from-scratch` repository, the `start-learning` skill serves as the deterministic entry point for the entire curriculum. It performs a one-time onboarding process that persists your learning configuration and establishes your baseline expertise level before you dive into AI engineering concepts.

## How to Invoke the start-learning Skill

The skill adapts to your host environment through a flexible **host invocation contract** defined in [`skills/start-learning/SKILL.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/skills/start-learning/SKILL.md)【L28-L35】.

### Codex-Style Hosts

On Codex-style platforms, invoke the skill as plain text:

```text
start-learning

```

### Claude-Code Hosts

On Claude-Code environments, use the slash command syntax:

```text
/start-learning

```

### Natural-Language Hosts

On any chat-based interface, you can trigger the skill conversationally:

```text
User: "I want to begin the AI engineering curriculum."
Agent: "Sure! Let's run the start-learning skill."

```

## The Onboarding Workflow

When invoked, the skill executes a deterministic six-step workflow that ensures consistent setup across all environments.

### Resume Routing Check

Before initiating new onboarding, the skill checks for existing state files including [`LEARNING.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/LEARNING.md) and [`MCP-LEARNING.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/MCP-LEARNING.md)【L41-L61】. If found, it routes you to the appropriate "learn" skill instead of duplicating work, preventing state corruption and preserving your progress history.

### The Three-Question Interview

The skill conducts a brief interview to capture your learning context:

1. **Why are you learning AI engineering?** - Captures your motivation
2. **How much time can you devote weekly?** - Establishes commitment level
3. **What do you most want to build?** - Defines your project goals

Your answers are recorded verbatim and embedded in the **Mission** section of the generated [`LEARNING.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/LEARNING.md) file【L05-L15】.

### Placement Quiz Execution

The skill invokes the companion **`find-your-level`** skill (defined in [`skills/find-your-level/SKILL.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/skills/find-your-level/SKILL.md)) to administer a 10-question assessment【L24-L34】. This quiz maps your existing knowledge to an entry phase between 0 and 19.

If you already know your desired start phase, you can skip the quiz and record the placement as `self-selected`.

### Generating Your Personalized Path

Using your interview responses, placement results, and the static roadmap from [`ROADMAP.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/ROADMAP.md), the skill generates a comprehensive [`LEARNING.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/LEARNING.md) file【L37-L69】 containing:

- **Mission** - Your verbatim interview answers
- **Placement** - Entry phase (0-19) and rationale
- **Path** - Table of all 20 phases with status indicators (`Skip`, `Do`, etc.) and estimated hours from the roadmap
- **Progress log** - Empty template for tracking completion
- **Review queue** - Spaced repetition tracking for concepts

## Understanding the LEARNING.md Output Structure

The generated file serves as your persistent learning state. Here is a simplified example of the output structure:

```markdown

## Mission

- **Why**: I want to build an autonomous agent
- **Time**: ~5 hours per week
- **Build**: An autonomous agent that can browse the web

## Placement

- **Entry Phase**: 7 ("Model Context Protocol")
- **Method**: Quiz completed

## Path

| Phase | Topic | Status | Est. Hours |
|-------|-------|--------|------------|
| 0-6 | Foundations | Skip | 0 |
| 7 | Model Context Protocol | Do | 8 |
| 8-19 | Advanced topics | Pending | 45 |

```

## Resuming vs. Restarting Your Learning Journey

### Resuming Existing Progress

To continue where you left off after the initial setup, use:

```text
learn

```

The host reads your existing [`LEARNING.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/LEARNING.md) and dispatches to the appropriate learning skill without re-running onboarding.

### Restarting with Confirmation

The skill is **idempotent** - it will never overwrite an existing [`LEARNING.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/LEARNING.md) without explicit confirmation【L93-L103】. To restart:

```text
> start-learning
Your existing LEARNING.md will be archived as LEARNING-2023-08-26.md.
Proceed with fresh onboarding? (yes/no) 
> yes

```

This archiving preserves your history while allowing a clean reset.

## Key Source Files

The following files implement the complete onboarding experience:

- **[`skills/start-learning/SKILL.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/skills/start-learning/SKILL.md)** - Core skill definition, host contracts, and file generation logic
- **[`skills/find-your-level/SKILL.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/skills/find-your-level/SKILL.md)** - Placement quiz implementation
- **[`ROADMAP.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/ROADMAP.md)** - Source of phase descriptions and estimated hours
- **[`learning-paths/model-context-protocol.json`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/learning-paths/model-context-protocol.json)** - MCP-specific learning path definitions
- **[`learning-paths/agent-skills.json`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/learning-paths/agent-skills.json)** - Agent skills track definitions

## Summary

- The `start-learning` skill creates a persistent [`LEARNING.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/LEARNING.md) file that serves as your curriculum compass
- It supports **three invocation styles**: plain text (Codex), slash commands (Claude-Code), and natural language
- The workflow includes **resume routing** to prevent duplicate onboarding and **idempotent** restart capabilities
- A **10-question placement quiz** maps you to phases 0-19, or you can self-select your entry point
- The generated file contains your **Mission**, **Placement**, **Path** (20 phases), **Progress log**, and **Review queue**
- Related files include [`ROADMAP.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/ROADMAP.md) for phase definitions and [`skills/find-your-level/SKILL.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/skills/find-your-level/SKILL.md) for assessment logic

## Frequently Asked Questions

### What happens if I already have a LEARNING.md file?

The skill checks for existing state files before running【L41-L61】. If found, it routes you to the `learn` skill instead of creating duplicates. You can force a restart, but the system will archive your existing file and request explicit confirmation before overwriting【L93-L103】.

### Can I skip the placement quiz?

Yes. If you already know which phase you want to start with, you can bypass the `find-your-level` quiz and record your placement as `self-selected`【L24-L34】. This is useful for experienced practitioners who want to jump directly to advanced topics like Model Context Protocol.

### Is it safe to run start-learning multiple times?

Yes, the skill is designed to be **idempotent**. It detects existing learning plans and either routes to the resume flow or asks for confirmation before archiving and restarting. You can safely invoke it from different agents or host environments without corrupting your learning state【L93-L103】.

### What's the difference between start-learning and learn?

**`start-learning`** performs the one-time onboarding interview, placement quiz, and initial [`LEARNING.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/LEARNING.md) generation. **`learn`** reads your existing [`LEARNING.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/LEARNING.md) and resumes active study from your current phase. Think of `start-learning` as initialization and `learn` as the runtime loop.