iOS QA agents go autonomous
An engineer built a ‘QA Engineer’ AI agent using Maestro that analyzes code, spins up simulator tests (clicks/navigates), self‑corrects with screenshots and coordinate estimation, and persists scripts in memory—demonstrating autonomous iOS test generation in the wild. Related efforts include an App Store Connect CLI agent skill for automated deployments and AI‑native Markdown test management products aimed at regulated QA/SRE workflows. (x.com) (x.com) (x.com)
Whitesmith published a detailed walkthrough on Feb. 23, 2026 that describes an agentic QA loop which captures screenshots and the iOS accessibility UI hierarchy and routes those observations to Claude Sonnet for decisioning. (whitesmith.co)) The Whitesmith post includes a code snippet showing the loop: get screenshot + UI hierarchy, append to LLM messages, call claude-sonnet-4-5, parse tool calls from the response, and act until TASK_COMPLETED or TASK_FAILED. (whitesmith.co)) A public GitHub skill named ios-dev from AlphaSquadTech exposes an agent skill that can build with xcodebuild, boot and manage iOS Simulators, take screenshots, record video, and navigate apps via Maestro as part of an autonomous build‑run‑verify loop. (github.com)) The ios-dev skill lists platform requirements—macOS with Xcode, Java 17+ (for Maestro), and Maestro itself (auto‑installed by the skill)—and offers slash commands for build, test, debug, and screenshot workflows. (github.com)) A separate, widely starred GitHub collection (rudrankriyam/app-store-connect-cli-skills) contains 56 commits and 448 stars and provides agent skills for automated TestFlight workflows, end‑to‑end release flows, and a repeatable simulator screenshot pipeline. (github.com)) Open-source projects aiming at AI‑native test management include Qualityfolio, a Markdown‑first framework with 71 commits that promises Git‑first test artifacts and SQL‑backed, audit‑grade evidence for regulated QA and SRE workflows. (github.com)) Commercial tools in the same space advertise scale and AI features—Testomat.io claims support for 100,000+ tests per project, while TestMu AI publishes case claims such as “70% faster test execution” for AI‑native test orchestration. (testomat.io)) Because the Simulator control and xcodebuild steps require macOS/Xcode, these autonomous agent skills are tied to macOS environments or macOS CI runners, which shapes where and how teams can run fully autonomous iOS QA agents. (github.com))
