How I Built an AI‑Powered iOS Automation Test Suite in One Afternoon (Step‑by‑Step Guide)

Solution Overview

Implementation combines three components:

CoPaw (or compatible multimodal model) – language model that can process screenshots (e.g., Kimi 2.6).

iOS simulator Skill – a skill that translates high‑level intents into simctl / idb commands for the iOS simulator.

iPhone 17 simulator on macOS – execution host; a real device can be substituted.

Environment Setup

1. Install iOS simulator Skill

Download the skill package from the Cocoloop skill store ( https://hub.cocoloop.cn/). If the domain https://dl.cocoloop.cn/ is unreachable, install the Molili agent ( https://www.molili.com.cn) and use its built‑in installer.

# Install Cocoloop manager
curl -L https://dl.cocoloop.cn/cocoloop-main.zip -o cocoloop.zip
unzip cocoloop.zip -d ~/.cocoloop

# Install iOS‑simulator skill
curl -L https://dl.cocoloop.cn/bss/skills/ios-simulator.zip -o ios-simulator.zip
unzip ios-simulator.zip -d ~/.cocoloop/skills

2. Install UI‑automation dependencies

macOS with Xcode Command Line Tools (install or update with

sudo rm -rf /Library/Developer/CommandLineTools && sudo xcode-select --install

)

Add Meta’s Homebrew tap: brew tap facebook/fb Install idb-companion and the Python library fb-idb:

brew install idb-companion
python3 -m pip install --upgrade fb-idb

Python 3.12 is recommended to avoid compatibility issues.

3. Install iOS simulator runtimes

Open Xcode → Preferences → Components and download the required iOS runtime images for the target device version.

iOS Simulator Skill Details

The skill wraps simctl and idb to provide:

Simulator lifecycle: create, boot, shutdown, erase, delete

App management: install, uninstall, launch, terminate

UI automation via accessibility tree: find, tap, type

Screenshot and screen‑recording

Privacy permission control

Push‑notification simulation

Basic usage examples:

# Health check
node skills/tristanmanchester-ios-simulator/scripts/ios-sim.mjs health

# List simulators
node skills/tristanmanchester-ios-simulator/scripts/ios-sim.mjs list

# Select and boot a default simulator (e.g., iPhone 16)
node skills/tristanmanchester-ios-simulator/scripts/ios-sim.mjs select --name "iPhone" --runtime "iOS" --boot

Automation Test Cases

Case 1 – Safari Baidu Search

Using the iOS‑simulator skill, open Safari on the iPhone 17 simulator, navigate to https://www.baidu.com, and search for “TestHub”. The AI agent consumes tokens to invoke the skill, open the URL, and perform the search. The process succeeds once the prompt is clear.

Case 2 – TestHub Smoke Test & H5 Compatibility

Goal: verify top‑menu navigation on https://testhub.aisky.cloud/ and ensure H5 pages render correctly. The AI initially missed the hamburger menu required to reveal hidden buttons; after refining the prompt a few times, it successfully traversed all menu items, captured screenshots, and produced a final test report.

Conclusion

The workflow demonstrates that AI‑assisted iOS automation is feasible with minimal manual setup. Alternatives such as writing a real‑device skill or converting the workflow to an Appium‑based skill are possible, but the core idea remains: combine a powerful language model, a multimodal perception layer, and a skill that translates high‑level intents into simulator commands.