AI Coding Task Brief Builder for Expo

Turn rough Expo and React Native app requests into safe, scoped, paste-ready task briefs for AI coding agents such as Codex, Claude Code, Cursor, and Windsurf.

What this Agent does

This Agent converts vague development requests into clear coding task briefs that include:

• Goal
• Scope for the task
• Target-file guidance
• Implementation requirements
• Do not change / cautions
• Verification steps
• Post-work report format
  It is designed to help AI coding agents work with a narrower scope, reduce unnecessary edits, and avoid accidental git operations.

Best for

• Expo app development
• React Native mobile apps
• Indie app MVPs
• UI fixes
• Bug fixes
• Feature additions
• Navigation changes
• AsyncStorage / persistence updates
• Push notification work
• APK and EAS Build preparation
• Android-focused app development

Key safeguards

Every generated task brief includes safeguards to:

• Preserve existing behavior unless the requested change requires otherwise
• Avoid unrelated refactoring
• Avoid unrelated file changes
• Avoid unnecessary dependency or package changes
• Prohibit git add, git commit, and git push

Language support

The Agent matches the user's input language whenever possible.

• English input → English task brief
• Japanese input → Japanese task brief
• Other languages → Same language when possible
• Unclear language → English by default
  Code, commands, file paths, package names, dependency names, and technical identifiers are kept unchanged.

Example input

The edit screen is not saving correctly.

## Example output

Title:
Fix edit screen save behavior

Goal:
Identify and fix the issue preventing the edit screen from saving changes correctly while preserving the existing edit flow and related app behavior.
Scope for this task:
Focus only on the edit screen save flow, the related state update logic, and persistence behavior. Do not redesign the screen or rewrite unrelated navigation.

Target files:
Inspect the existing project structure and decide which files are directly related to the edit screen, save handler, navigation, and persistence.
Implementation requirements:
1. Identify the likely cause before making changes.
2. Verify whether the issue is caused by form state, route params, save handler logic, AsyncStorage persistence, or navigation after saving.
3. Apply the smallest necessary fix.
4. Preserve the existing UI unless a UI change is required for the fix.
5. Ensure saved edits are reflected when returning to the relevant list or detail screen.

Do not change / cautions:
- Do not change unrelated files.
- Do not perform unrelated refactoring.
- Preserve existing behavior unless the requested fix requires otherwise.
- Do not change package versions or dependencies unless necessary.
- Do not run git add, git commit, or git push.

Verification steps:
- Run `npx expo start`.
- Test the edit screen on an Android emulator or physical Android device.
- Confirm that editing an existing item saves correctly.
- Confirm that the updated data appears after navigating back.
- Run `npm run lint` if available.

Post-work report format:
1. Modified files
2. Added files
3. Deleted files
4. Implementation summary
5. Verification performed
6. Unverified items / cautions
7. What the user should check next
After completing the work, provide a work report for ChatGPT review. Do not run git add, git commit, or git push.

## Example prompts

Try requests like:
1)Fix the home screen layout on small Android screens.

2)Add a bottom tab bar with Home, Record, Logs, and Settings.

3)Prepare my Expo app for an Android APK build.

4)通知設定画面を追加したい。

##Why use it

AI coding agents are powerful, but vague prompts can lead to broad edits, unnecessary refactoring, or changes outside the intended scope.

This Agent helps you give the AI a safer task brief before asking it to code.