Skip to content

Tool Guides

Salman Ashraf edited this page May 29, 2026 · 1 revision

Tool Guides

How to use Mobile Dev Skill Agents with Claude Code, Cursor, ChatGPT, and GitHub Copilot.


Claude Code

How Agents Load

CLAUDE.md at the repo root is read automatically whenever Claude Code is opened in this directory. It contains the full agent index, severity level definitions, and platform conventions — Claude already knows the agents without any setup.

Running an Agent

# Reference an agent directly:
Use the agent at agents/android/android-crash-analyzer/agent.md to analyze this crash:
[paste crash log]
# Reference a file:
Use the agent at agents/android/code-reviewer/agent.md to review:
@examples/android/ProfileViewModel.kt

Multi-File Review

@ProfileViewModel.kt @ProfileRepository.kt @ProfileUseCase.kt
Review the full data flow for Clean Architecture violations.

Full guide: docs/getting-started.md


Cursor

How Agents Load

.cursorrules at the repo root is read automatically in every Cursor session. It contains the full agent index, platform conventions, and severity levels. Cursor Chat already knows the agents.

Cursor Chat (single file review)

@agents/android/code-reviewer/agent.md
Review this ViewModel for Clean Architecture violations:
[paste code]

Cursor Composer (apply fixes automatically)

Open Composer (Cmd+I), then:

@agents/android/android-crash-analyzer/agent.md @crash.txt
Analyze this crash and apply the recommended fix to the affected file.

Composer shows you a diff of every change — accept or reject individually.

Multi-File Review

@agents/android/code-reviewer/agent.md
@ProfileViewModel.kt @ProfileRepository.kt
Review both files together and flag any layer boundary violations.

Full guide: docs/cursor-integration.md


ChatGPT

Method 1 — Manual Paste (any tier, no setup)

  1. Open agents/<platform>/<name>/agent.md
  2. Copy the ## System Prompt block
  3. Paste as your first message in a new chat — GPT acknowledges the role
  4. Follow with your code using the input format from ## Input Format

Method 2 — Custom GPT (recommended for teams)

A ready-made Custom GPT system prompt lives in docs/chatgpt-integration.md. It routes to the right agent automatically based on what you describe — one GPT covers all 16 agents.

To set up:

  1. Go to ChatGPT → My GPTs → Create a GPT → Configure tab
  2. Paste the system prompt from docs/chatgpt-integration.md
  3. Save — then share the link with your team

Usage:

Review this Android ViewModel: [paste code]

The GPT detects the platform and applies the correct agent automatically.

Method 3 — Assistants API

import openai
client = openai.OpenAI()

assistant = client.beta.assistants.create(
    name="Android Code Reviewer",
    instructions=open("agents/android/code-reviewer/agent.md").read(),
    model="gpt-4o",
)

Full guide: docs/chatgpt-integration.md


GitHub Copilot (VS Code)

How Agents Load

.github/copilot-instructions.md is read when you use @workspace in Copilot Chat. It contains condensed agent rules for all 16 agents.

Copilot Chat

@workspace Review this file using the Android Code Reviewer rules:
#file:ProfileViewModel.kt
@workspace Analyze this crash log using the Android Crash Analyzer rules and return
the 9-section report: [paste crash log]

Copilot Edits (inline fix)

Select the problematic code block → Cmd+I (Mac) or Ctrl+I (Windows):

Fix the GlobalScope leak and remove the direct UserRepository instantiation

Copilot Edits proposes a diff inline — accept or discard each change.

@workspace for Architecture Questions

@workspace Using Clean Architecture rules for Android, does this project correctly
separate domain, data, and presentation layers? List any violations.

Full guide: docs/vscode-copilot-integration.md


Comparison

Claude Code Cursor ChatGPT Copilot
Auto-loads agents Yes (CLAUDE.md) Yes (.cursorrules) No — paste manually or use Custom GPT Partial (@workspace)
Apply fixes directly Yes Yes (Composer) No Yes (Copilot Edits)
Multi-file context Yes Yes (@file) No (paste manually) Yes (@workspace)
Best for Full agent sessions Daily review + fix Team sharing, API Inline quick fixes

Home · Using Agents · Platform Guide · Skills and Prompts