Starting a mobile app with Claude Code or Cursor: why start from a kit, not a blank project
An AI coding agent is only as good as the project it opens. Ask Claude Code or Cursor to build a mobile app in an empty create-expo-app folder and you get one of two results. Sometimes the first screen works in one prompt. More often the agent invents its own styling pattern, installs a second date picker, and leaves you with code you cannot explain a week later. The model is rarely the cause.…
An AI coding agent's performance heavily depends on the project it opens. If you ask Claude Code or Cursor to build a mobile app in an empty create-expo-app folder, you may encounter inconsistent styling, extra components, and code that is difficult to understand later. The problem doesn't lie in the model itself; it's due to the empty project providing no guidelines, leading the agent to make varying decisions on each prompt.
This article discusses what a starting kit changes, what the agent needs prior to writing any code, and situations where a kit is not helpful.
Four recurring failure modes emerge when working with a blank project: lack of conventions, dependency drift, delayed theme support, and inconsistent design. Fixing these issues individually is manageable, but collectively, they transform a weekend prototype into a maintenance nightmare.
A kit offers the agent three critical elements it cannot obtain from a prompt: existing examples of various patterns, pre-existing components, and a written project instruction file. The former provides working patterns, the latter reduces the effort required to create new components, and the file serves as a contract that outlines how the codebase operates.
The file, read by both Claude Code and Cursor, serves as the project's instruction manual. It outlines the stack, including specific versions, and outlines the rules governing the codebase. For instance, hex color codes are not hard-coded outside of a designated theme file, and toggles always utilize a specific library. The file also prevents the inclusion of features that are outside the kit's scope.
An essential aspect of the file is its focus on the agent rather than the reader. It outlines the rules that a new engineer would need on their first day, ensuring consistency in the agent's work. The file changes how the agent approaches various features, such as handling sensitive operations (like payments) on the server side and data hooks.
Moreover, the file flags potential pitfalls, like the need for additional configuration variables in specific routes, and provides guidance on task recipes. It links to a repository of tested prompts, tailored to the kit's specific file structure, ensuring the agent starts from familiar names rather than generic ones.
In practice, the rules in the file significantly impact the agent's behavior. For example, it determines whether a feature should be implemented on the client-side or server-side, guides data handling, and even highlights critical aspects that must be considered when writing hooks or queries. By incorporating these guidelines, the agent has a consistent framework for creating the app, reducing the likelihood of errors and ensuring a more polished final product.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.