Skip to content
Salman Ashraf edited this page Oct 22, 2026 · 11 revisions

Mobile AI Agents

A mobile engineering system of AI agents, reusable skills, and end-to-end workflows for Android, iOS, Flutter, React Native, Kotlin Multiplatform, Unity, and Unreal teams.

Mobile AI Agents helps you plan, build, review, test, verify, and release mobile apps with repeatable AI workflows instead of one-off prompts.


What You Can Use

Resource Count Best For
Agents 36 Structured specialist work such as crash analysis, code review, UI generation, security scanning, release notes, and store copy
Skills 63 Focused prompt modules you can compose into any coding session
Workflows 17 End-to-end loops such as issue-to-agent, device proof reports, mobile flight recorder, mobile harness, release cycle, and app launch

Start here:

  • Mobile Engineering Loop - see how planning, implementation, verification, reporting, and saved context fit together.
  • Agents - choose the right specialist agent and see how to invoke it.
  • Skills and Prompts - browse every skill by platform and use case.
  • Android Device QA Setup - check Android tooling and configure Mobile MCP for /mobile-mcp-qa.
  • Mobile App Design - create, redesign, and reskin mobile app UI.
  • Mobile Harness - guide a first app from stack selection through implementation and device proof.
  • Enterprise Mobile Harness - govern runs with risk-based policy, scoped changes, rollback, and auditable evidence.
  • Workflows - choose a full process when you need planning, QA, release, or device proof.
  • Tool Guides - use the toolkit in Claude Code, Cursor, ChatGPT, GitHub Copilot, Windsurf, and Codex.

Quick Start

Install into your AI coding tool:

npx mobile-ai-agents install

Common options:

npx mobile-ai-agents install --tool claude
npx mobile-ai-agents install --tool cursor
npx mobile-ai-agents install --tool windsurf
npx mobile-ai-agents install --tool copilot
npx mobile-ai-agents install --tool codex
npx mobile-ai-agents install --tool all

Then use a specialist:

@AXIOM review this Android ViewModel for Clean Architecture issues.

/mobile-mcp-qa verify this Android flow on an emulator.

device-proof-report turn the Mobile MCP screenshots and actions into DEVICE_QA_REPORT.md.

Prepare Android device QA from the app project:

npx mobile-ai-agents doctor --platform android
npx mobile-ai-agents mcp setup --client claude
npx mobile-ai-agents mcp config init --app-id com.example.app

Initialize governed Mobile Harness runs:

npx mobile-ai-agents@1.0.37 harness policy init --profile enterprise
npx mobile-ai-agents@1.0.37 harness policy validate

Common Problems And Where To Start

I need to... Start With
Review Android Kotlin or Compose code @AXIOM, code-reviewer, /code-review, /compose-review
Investigate an ANR anr-investigation, @FREEZE, /anr-investigation
Investigate a memory leak memory-leak-analyzer, @RETAINER, /memory-leak-investigation
Debug a crash @CRASHER, android-crash-analyzer, crash-analyzer-ios, /crash-triage
Fix startup, jank, ANR, memory, battery, network, or app size @PERF, @FREEZE, @RETAINER, /perf-audit, perf-sprint
Review Swift or SwiftUI code swift-reviewer, @SWIFT, /swift-review, /swiftui-review
Build a Flutter feature bloc-feature-builder, @DART, /flutter-review, /flutter-tdd
Optimize React Native performance-optimizer, @BRIDGE, /rn-performance, /bridge-audit
Verify an app on a device Android Device QA Setup, /mobile-mcp-qa, device-proof-report
Preserve project context for the next AI session mobile-flight-recorder, Mobile Memory
Run release gates app-launch, release-cycle, /security-audit, /release-prep
Turn a GitHub issue into a plan issue-to-agent
Create or reskin mobile UI /mobile-app-design
Build a first app and choose a stack MOBILE-HARNESS in BEGINNER_GUIDED mode
Govern AI changes for a larger organization Enterprise Mobile Harness

Device Proof Reports

Device Proof Reports answer a simple question:

Did this mobile build actually run correctly on a real target, and where is the evidence?

Run /mobile-mcp-qa first to install or launch the app, tap through the flow, and capture screenshots. Then run device-proof-report to create DEVICE_QA_REPORT.md with:

  • device, OS, app id, build, commit, and orientation
  • actions performed through Mobile MCP
  • screenshots and what each screenshot proves
  • assertions marked PASS, FAIL, or BLOCKED
  • crashes, logs, accessibility notes, and performance notes
  • bugs with repro steps and next fixes

Use this for PR handoff, release gates, bug fixes, screenshot validation, and Mobile Harness reports.


Mobile Flight Recorder

Mobile Flight Recorder preserves project context so any AI tool can resume days or weeks later without asking the user to re-explain the app.

Run it at the end of every meaningful session:

Run Mobile Flight Recorder. Update MOBILE_AGENCY_CONTEXT.md and MOBILE_MEMORY.md if useful. Capture decisions, files changed, tests run, known bugs, device setup, links, and exactly one Next Recommended Action. Do not include secrets.

It records app goal, platform, stack, architecture, implemented features, current task, pending tasks, known bugs, real test commands, design source, device/Mobile MCP setup, decisions, changed files, evidence, links, and one next recommended action.

Resume any AI tool with:

Read MOBILE_AGENCY_CONTEXT.md and MOBILE_MEMORY.md if present. Continue from Next Recommended Action. Inspect changed files before editing and update the flight recorder before stopping.

Performance Loop

Use the Performance Loop when users report slowness, freezes, jank, memory growth, battery drain, network waste, or app size problems. Start with evidence, not guesses:

1. Identify the symptom: startup, ANR, memory, frame, battery, network, size, slow screen.
2. Collect device, OS, app version, commit, flow, profiler traces, screenshots, logs, and baseline numbers.
3. Route to @PERF, @FREEZE, @RETAINER, /perf-audit, /anr-investigation, or /memory-leak-investigation.
4. Fix one bottleneck at a time.
5. Re-measure the same flow on the same target.
6. Produce PERFORMANCE_AUDIT_REPORT.md with before/after metrics and regression monitoring.

Use this for release gates, slow-screen bugs, ANR investigations, memory leak reports, and any PR where performance risk matters.


Wiki Navigation

Page What It Covers
Agents All 36 agents, grouped by platform, with usage guidance
Skills and Prompts All 53 skills and the prompt library
Android Device QA Setup Android doctor, Mobile MCP client setup, project flow config, and troubleshooting
Mobile App Design How to use /mobile-app-design for new screens, redesigns, and app reskins
Mobile Harness PRD-to-device verified implementation loop
Enterprise Mobile Harness Policy, risk, scope, gates, rollback, evidence, schemas, and CI integration
Workflows All 16 workflows and when to use each
Using Agents How to choose, apply, and interpret any agent
Tool Guides Claude Code, Cursor, ChatGPT, GitHub Copilot, Windsurf, Codex
Platform Guide Per-platform tips and gotchas
Reference Input fields and output formats
Contributing How to build and submit a new agent, skill, or workflow
FAQ Common questions

Repository: github.com/salmanashraf/mobile-ai-agents

Complete Documentation

Clone this wiki locally