AI coding assistants
Choose a tool that can read the folder, edit files, and run commands.
The best tool for this kit is not the one with the fanciest chat window. It is
the one that can open the whole project, read AGENTS.md, run terminal commands,
show a diff, and keep working through errors.
Good fits
| Style | Examples | Best for |
|---|---|---|
| Coding agent | Codex, Claude Code, OpenCode | Multi-file changes, commands, tests, long tasks |
| AI editor | Cursor, VS Code with Copilot/Claude/Codex/Gemini | Editing while you watch files and diffs |
| Web chat | ChatGPT, Claude, Gemini | Planning and explanations; slower for direct project work |
Web chat can help, but you will need to paste files and apply changes yourself. For this starter, a tool with direct folder access is much easier.
First prompt
Read AGENTS.md before changing anything. Then read START-HERE.md, DESIGN.md,
DESIGN-DARK.md, and FEATURES.md. Summarize the app, tell me which parts are mock,
and get it running without wiring external services.Prompts that work well
Be clear about the outcome and constraints:
Turn Vela into a dog-training coach. Keep the four-tab structure, mock mode,
light/dark themes, English/Spanish, accessibility behavior, and every integration
seam. Ask me for missing product decisions before changing provider setup.Connect Supabase authentication using the production adapter already in the
repo. Read the matching agent skills first. Do not put server secrets in the app.
Stop before any paid account, deployment, or store submission action.Keep approvals meaningful
It is normal for the agent to ask before signing into an account, creating a cloud project, changing a team-owned EAS project, deploying, purchasing a plan, or submitting to a store. Those actions affect systems outside your folder.