Vibe Coding · Leçon

Découper les grandes tâches en étapes

Décomposez une fonctionnalité en éléments pouvant faire l’objet de prompts.

Découper les grandes tâches en étapes est une leçon Vibe Coding gratuite sur CoddyKit. Ceci est la leçon 3 sur 4. Tu peux lire la leçon complète ci-dessous gratuitement — puis la pratiquer en direct dans le navigateur avec un éditeur de code intégré et un tuteur IA 24/7. Elle fait partie du parcours d'apprentissage Vibe Coding, et ta progression se synchronise sur le web et l'application CoddyKit. Le cours Vibe Coding comprend 4 leçons au total.

Certaines parties de cette leçon n'ont pas encore été traduites et s'affichent en anglais.

Big Tasks Break AI

Ask AI to "build me a todo app with login, a database, and dark mode" in one prompt and you'll often get a tangled mess — half-finished, buggy, impossible to review. AI, like people, does its best work on one clear thing at a time.

The skill in this lesson is decomposition: breaking a big feature into small, promptable chunks you build and verify one by one.

Why Small Steps Win

Building in small steps with AI gives you huge advantages:

Reviewable — you can actually read 20 lines, not 500.
Testable — run each piece and confirm it works before moving on.
Recoverable — if step 3 breaks, steps 1–2 are still safe.
Steerable — you correct course early, not after a giant mess.

Small steps are how pros ship reliable apps with AI.

Plan First, Then Prompt

Before writing code, have the AI help you plan. Ask it to break your feature into a numbered list of steps. You're not building yet — you're getting a roadmap you can follow and reorder.

This planning prompt costs one message and saves you from building the wrong thing.

You are a senior engineer. Don't write code yet. I want to build a 'save favorite articles' feature for my web app. Break this into a numbered list of small, buildable steps, from simplest to most complete. Keep each step to one clear task.

A Good Decomposition

A solid plan turns one scary feature into a checklist. For our "favorites" feature, the AI might return:

1. Add a star button to each article card.
2. Toggle the star's filled/empty state on click.
3. Store the favorited IDs in the browser.
4. Load saved favorites when the page opens.
5. Add a "My Favorites" view that filters the list.

Now each line is its own small prompt.

Prompt One Step at a Time

With the plan in hand, build step 1 only. Tell the AI explicitly to stop there. This keeps the output small and focused.

Notice the constraint: only this step. That prevents the AI from racing ahead and writing the whole feature in one unreviewable blob.

We're following our plan. Build ONLY step 1: add a star button to each article card. Use vanilla JS and HTML. Don't implement saving or toggling yet — just render the button. Return only the code for this step.

Verify Before You Continue

After each step, run it and check. Does the button appear? Good — move on. Broken? Fix this step before adding more. Stacking new code on a broken foundation is how vibe-coded apps fall apart.

Here's a tiny, self-contained piece of step 2 logic — toggling a saved state — that you could verify on its own.

// Step 2 logic, tested in isolation
function toggleFavorite(savedIds, id) {
  if (savedIds.includes(id)) {
    return savedIds.filter((x) => x !== id);
  }
  return [...savedIds, id];
}

let ids = [];
ids = toggleFavorite(ids, 7);
console.log(ids);
ids = toggleFavorite(ids, 7);
console.log(ids);

Carry Context Between Steps

Each new step depends on the last. When you prompt for step 2, remind the AI what step 1 produced — paste the relevant code or describe it. In agentic tools like Claude Code or Cursor, the AI can see your files, so just point at them.

Without this context, the AI may invent different names and break the connection between steps.

Now build step 2: toggle the star when clicked. Here is the button from step 1 so you match its class and structure:

<button class="star-btn" data-id="42">☆</button>

Write the click handler that switches between ☆ and ★. Return only the JS for this step.

The Vertical Slice Approach

Another way to decompose: build one tiny end-to-end slice first, then widen it. Instead of building the whole UI, then the whole backend, make ONE feature work all the way through — then add the rest.

Slice 1: favorite a single hard-coded article, save it, reload it.
Then: apply that working flow to all articles.

A thin working slice beats many half-built layers.

Let's build the thinnest working version first. Make ONLY this work end-to-end: clicking a star on one article saves its id to localStorage and re-reading localStorage shows it's saved. Skip the UI polish and the favorites page for now.

Decomposing in App Builders

Even in prompt-to-app tools like Bolt, v0 and Lovable, small steps win. Generate the first screen, preview it, then ask for the next change. Don't request the entire app in your opening prompt.

Prompt 1: "A landing page with a hero and a sign-up button."
Prompt 2: "Now add a pricing section below the hero."
Prompt 3: "Make the sign-up button open a modal."

You build a real app through a conversation, not one giant ask.

Keep a Running Checklist

Track your steps so you and the AI stay aligned. Keep a simple checklist in your prompt or a notes file and mark progress. This becomes the shared memory of your build.

In Claude Code, a TODO.md or CLAUDE.md works great — the agent reads it and knows what's done and what's next.

Here is our progress. Continue from the first unchecked item.

[x] 1. Star button on each card
[x] 2. Toggle star on click
[ ] 3. Save favorited IDs to localStorage
[ ] 4. Load favorites on page open
[ ] 5. 'My Favorites' filter view

Build step 3 only.

Knowing When to Split Further

If the AI's output for one "step" is still large or buggy, that step was too big — split it again. Signs a step needs breaking down:

The answer is hundreds of lines.
It touches many files at once.
You can't tell if it works by reading it.

There's no shame in tiny steps. The smaller the step, the more control you keep.

Quick Check

You ask AI to build a whole feature at once and get 400 lines you can't follow. What's the best next move?

Recap: Build in Steps

Decomposition is how you tame big builds with AI:

Have the AI plan the feature as a numbered list first.
Prompt one step at a time and tell it to stop there.
Run and verify each step before moving on.
Carry context forward and keep a checklist.
Prefer a thin end-to-end slice, and split further when a step gets too big.

Next: pinning the exact tech stack and style so AI builds it your way.

Questions Fréquemment Posées

La leçon « Découper les grandes tâches en étapes » est-elle gratuite ?

Oui — le texte complet de « Découper les grandes tâches en étapes » est gratuit à lire ici sur le web. Pour la pratiquer de manière interactive (un éditeur de code intégré et un tuteur IA 24/7) et déverrouiller le reste du cours Vibe Coding, passe à CoddyKit PRO. Le cours Vibe Coding comprend 4 leçons au total.

Qu'est-ce que j'apprendrai dans « Découper les grandes tâches en étapes » ?

Décomposez une fonctionnalité en éléments pouvant faire l’objet de prompts. Tu pratiques Vibe Coding avec du code pratique que tu exécutes directement dans le navigateur, et un tuteur IA 24/7 répond à tes questions au fur et à mesure que tu avances dans la leçon.

Dois-je avoir de l'expérience pour commencer Vibe Coding ?

Aucune expérience préalable n'est requise. Vibe Coding sur CoddyKit est structuré pour les débutants jusqu'aux apprenants avancés, donc tu peux commencer ici ou depuis le début et avancer à ton rythme. Ceci est la leçon 3 sur 4.

Combien de temps prend la leçon « Découper les grandes tâches en étapes » ?

La plupart des leçons CoddyKit prennent environ 5–10 minutes. Chacune est courte et interactive, tu progresses régulièrement et tu repiques exactement où tu t'es arrêté sur le web et l'app.

Peux-tu écrire et exécuter du code dans cette leçon Vibe Coding ?

Oui. Chaque leçon Vibe Coding inclut un éditeur de code intégré, tu écris et exécutes du vrai code directement dans ton navigateur et tu reçois des retours IA instantanés — aucune configuration locale requise.

Toutes les leçons de ce cours

← Retour à Vibe Coding