When should you use it?
- You keep a set of recipes you like but still stare at the fridge every evening
- You want weeknight dinners that fit a time limit
- You want to try a coding agent on something low-stakes and useful
Not for: Medical or clinical diets, which need a dietitian; People who want brand-new recipes every week rather than their own
What do you need, and what do you get?
You need
- Claude Code in the Claude desktop app
- A folder of recipes (the author offers a starter folder) with a short README
- Your family's dislikes, allergies, time limits and leftover habits
You get
- A FAMILY_PREFERENCES.md file the agent reads every time
- A seven-night dinner plan without repeats and with at least two leftover nights
- A shopping list grouped by produce, dairy, pantry, frozen, meat and seafood, and other, with quantities
How does it work, step by step?
- Set up the folderPut your recipes in one folder, or start from the author's sample, and open it in Claude Code.
- Paste the promptPaste the prompt below and approve the short plan the agent shows before it creates any files.
- Answer the questionsAnswer its questions about dislikes, time limits and leftovers; it saves them to a preferences file.
- Check the sample weekRead the sample plan and list, and correct anything that does not fit your family.
- Reuse it weeklyEach week, ask for a new plan; add recipes to the folder as you find ones you like.
The prompt
The original prompt by Marissa Goldberg, word for word, with a link to where it was published. Below it is our adapted version with fill-in blanks, if you want it. Works in Claude Code and Claude Cowork. Values in lime are placeholders.
Original prompt by Marissa Goldberg
You are my meal planning assistant. You have access to the local folder for the “cooking-with-claude” repo I just opened. Goal: Create a weekly meal plan (7 dinners) and a categorized shopping list based on my preferences. Rules: - Use meals from this repo first. If you need to suggest a new meal, ask me before adding anything new. - Avoid repeats in the same week. - Prefer weeknight-friendly meals unless I specify otherwise. - Include at least 2 leftover-friendly dinners. - When you generate a shopping list, group it by category (produce, dairy, pantry, frozen, meat/seafood, other) and include quantities. First-time setup: 1) Read the README to understand how this repo is organized. 2) Ask me up to 7 quick questions to capture my family preferences. 3) Save my preferences into a markdown file named FAMILY_PREFERENCES.md. 4) Confirm you’re ready by generating a sample 7-night plan + shopping list. Before creating or editing any files, show me a short plan of what you will do.
Exactly as published by Marissa Goldberg in AI Recipe: Build Your Personal Meal Planning Assistant.
Our adapted version (optional)
Fill in the blanks (optional; the copied prompt uses your values)
Help me plan dinners from the recipe files in this folder. Read FAMILY_PREFERENCES.md first. If it does not exist, ask me up to {max_questions} questions about dislikes, allergies, cooking time and leftovers, then save my answers there. Plan {nights} dinners for {week_of}: no dish twice, weeknights under {weeknight_minutes} minutes, at least {leftover_nights} that make good leftovers, and stay under {budget} if you can estimate it. Then write one shopping list grouped by store section with quantities for {servings} people, and mark items that are usually already in a pantry so I can check. Use only recipes in this folder unless I agree to a new one, and tell me your plan before you create or change any file.
What can go wrong?
- Allergy rules are only as good as the recipe files; check ingredient lists yourself for anything serious.
- Quantities are estimates; check them against what you already have before you shop.
- Keep the agent inside the recipe folder so it cannot touch other files.
- A new meal suggestion is only added after you agree; say no if you want to stay with known recipes.
Our take
A charming, practical first project for anyone curious about coding agents but not about code. The prompt is short and well judged: use my recipes first, ask before adding anything new, no repeats, weeknight-friendly, two leftover nights, a grouped list with quantities, and show a plan before creating or editing files. That last line is a good habit for any agent work. We ran it word for word in a coding agent on a fictional folder of twelve recipes, answering its questions for a family of four with a mushroom-averse child, a 35-minute weeknight limit, one vegetarian night and no shellfish. It showed its plan, asked seven questions, saved the preferences file and produced a seven-night plan that met every rule, with chili and lentil soup marked for leftovers and a list grouped by category. It did not invent new dishes. Check the quantities against your pantry before you shop.
Where the time goes back: Deciding dinners and writing the list is a weekly chore; this turns it into reviewing a plan the agent drafted from food you already know works.
FAQ
Do I need to code?
Where do the recipes come from?
What does the preferences file do?
Source and credit
The original prompt on this page is by Marissa Goldberg, shown word for word with a link to where it was published. The summary, notes and adapted version are ours.
- Original
- AI Recipe: Build Your Personal Meal Planning Assistant
- Creator
- Marissa Goldberg
- Published on
- Idea Kitchen (Substack)
- Original date
- Jan 22, 2026
- License
- Not openly licensed; the original prompt is shown word for word with credit and a link to the original
- Originally for
- Claude Code
- AI dry run
- Oct 22, 2026: passed. Dry-run tested by our AI test agent on sample input. Quantities are estimates; noted in pitfalls.
AI dry run on sample data; not run with live connected accounts. The playbooks.bot test agent filled the prompt with synthetic inputs, produced the output and checked it against the page's stated outputs and guardrails. - Verified
- Oct 23, 2026: source link resolves to the original, the credited creator published it, our excerpt or prompt matches it, and the license allows our use.
Are you the creator? Claim or correct this page.