What a coding agent does in a React Native app project
In a mobile project the agent takes the large, well-specified middle of the work. It scaffolds screens from agreed designs, wires up navigation stacks and tabs, builds list views for long feeds, writes form validation that applies the same rules as the server, and adds the loading, empty and error states that tend to be skipped late in a sprint. It writes unit and component tests next to the code, drafts store listing text for an engineer to edit, and carries out refactors that touch hundreds of files, such as moving every network call behind one typed client. We use Claude Code and Cursor. The agent works inside the real repository: it reads the existing code before changing it, plans, implements, then runs the type checker, linter and test suite before it proposes anything. It does not decide what the product should do. It works from a scoped task that an engineer wrote.
