Ingredient-based search.
You have three things in the fridge and a question: what can I make? Ask the internet and you get a recipe site that opens with a thousand-word childhood memoir, four ads, a newsletter popup, and the ingredient list somewhere near the bottom. What you wanted was ingredients in, meals out. This blueprint builds you a recipe finder that works the way you think about dinner: you describe it, the agent builds it, and it runs as your own app with nothing padding the page.
A recipe search app you drive by ingredients: type in what you have and filter the results down to meals you can actually cook. Results come back in a clean grid, each card showing calorie counts and cooking time up front, so you can weigh "fast" against "light" without opening every card. The filtering runs instantly as you adjust ingredients, so narrowing down feels like sorting a pile, not filling a form.
You watch it happen in the task workspace: the agent scaffolds the project, wires the ingredient filter, and confirms the grid renders and filters respond before handing you a live preview. The code lands on a branch in your repository with a pull request ready.
The first version is a base you steer with plain follow-up messages:
Each follow-up wakes the same repository and the agent continues on its own code.
Do I need to know how to code? No. You describe the finder you want; the agent writes it, verifies the search filters, and ships the code to your repository.
Where do the recipes come from? From a data source you control, seeded into your own app. Expand the list or wire in a recipe API whenever you want.
Can I add my own recipes? Yes. Ask for an add-recipe form and the agent wires it in, so your own dishes sit right alongside the rest.