React Native Audit

Version updated for https://github.com/maheshwarimrinal/react-native-agents to version v1.0.2.

  • This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

React Native Agents provides AI guidance to React Native developers. It identifies issues and suggests fixes by analyzing the project context, using deep references, and explaining findings with practical examples. The action can be used to automate pull-request reviews and is compatible with various AI coding tools like Claude Code, Cursor, Windsurf, GitHub Copilot, Codex, Zed, Aider, MCP clients, and GitHub Actions.

What’s Changed

React Native Agents v1.0.2

React Native Agents v1.0.2 — New Build, Doctor, Native Modules & Bundle Size Agents

Release notes

React Native Agents v1.0.2

This release expands React Native Agents with new implementation, diagnostics, native-module, and bundle-size capabilities.

Highlights

New specialist agents

  • rn-build — Create React Native screens, components, forms, lists, and data flows.
  • rn-doctor — Diagnose Android, iOS, Metro, dependency, environment, and build failures.
  • rn-native-modules — Review TurboModules, Fabric, JSI, Codegen, threading, and bridge-based native modules.

Deterministic bundle-size analysis

Added bundle analysis without an API key or model call:

npx @maheshwarimrinal/react-native-agents size

Compare bundle size against a base branch:

npx @maheshwarimrinal/react-native-agents size --base main

Budget checks are supported:

npx @maheshwarimrinal/react-native-agents size \
  --base main \
  --budget-delta 100kb

The analyzer now includes:

  • Source-map-based package attribution
  • npm, Yarn, pnpm, and Bun lockfile detection
  • Strict budget validation
  • Safe failure when the base bundle cannot be built
  • No unexpected CLI downloads through npx --no-install

Improved GitHub Action reliability

The GitHub Action now:

  • Fails by default when an agent cannot complete
  • Reports failed agents through action outputs
  • Detects incomplete reviews caused by budget exhaustion
  • Supports fail-on-error: false for intentionally partial reviews
  • Falls back to the GitHub files API when a pull-request diff is too large
  • Handles configuration files outside the repository root more reliably

Example:

- name: React Native audit
  uses: maheshwarimrinal/react-native-agents@v1.0.2
  with:
    provider: openai
    model: gpt-5
    api-key: ${{ secrets.OPENAI_API_KEY }}

To allow partial results:

with:
  fail-on-error: false

Use this only when a review is allowed to continue despite incomplete agent coverage.

Improved routing

MCP and GitHub Action routing now better identifies:

  • Build and environment failures
  • New screen and component requests
  • Native module and platform-code changes
  • Performance issues described without explicit React Native keywords

Interactive agents are excluded from automatic pull-request reviews unless explicitly requested.

Demo and documentation improvements

  • Added a real-world React Native audit demo project.
  • Added examples for Android native code and catalogue performance issues.
  • Split the README into focused documentation pages.
  • Updated generated outputs for Claude Code, Cursor, Windsurf, Copilot, and AGENTS.md.
  • Expanded evaluation coverage for build, doctor, and native-module scenarios.

Validation

This release includes expanded automated coverage for:

  • Agent routing
  • Bundle-size analysis
  • Source-map attribution
  • Budget enforcement
  • API failure handling
  • Build diagnostics
  • Native-module reviews
  • Generated target synchronization

Upgrade

npm install @maheshwarimrinal/react-native-agents@1.0.2

Or run the latest CLI directly:

npx @maheshwarimrinal/react-native-agents@1.0.2 install

Full comparison: https://github.com/maheshwarimrinal/react-native-agents/compare/v1.0.1...v1.0.2

One important release-note detail: mention that API and budget failures now fail the GitHub Action by default. This is a behavior improvement, but users who relied on partial reviews should configure:

fail-on-error: false