EAS Cloud Simulator
Run, control, and capture Expo apps remotely when a local iOS simulator or Android emulator is unavailable.
The documentation discloses paid usage, EXPO_TOKEN, remote daemon tokens, billing termination, dotenv cleanup, and gitignore requirements, and includes a macOS confirmation rule plus reset procedures; however, non-macOS use proceeds directly into a paid service, npx@latest, remote artifact downloads, and broad shell-tool allowances are not integrity- or consent-gated enough, so points are deducted.
The core flow includes platform branches, readiness polling, timeouts, retry limits, and concrete troubleshooting, with stated version ranges; however, it depends on experimental CLIs, the mutable latest tag, external services, and no committed test suite covering key paths. Static review cannot reproduce execution, so the score stays within the 10-point static ceiling and loses determinism and environment-coverage points.
Triggers, local-versus-cloud boundaries, Mac/non-Mac cases, and static-versus-development build modes are clearly described; however, Windows is deferred to WSL/Git Bash, Chinese-language interaction and mainland-China reachability are not addressed, and core functionality depends on Expo/EAS cloud services and public tunneling.
The skill provides frontmatter, use cases, prerequisites, mode selection, command tables, reference documents, troubleshooting, and limitations, while the skill and repository declare MIT; however, maintenance ownership, update responsibility, changelog, and stable-version policy are unclear, and latest plus experimental APIs create documentation-drift risk.
It supplies directly reusable sequences from session startup and builds through installation, connection, interaction, screenshots, and cleanup, and clearly separates live development from static builds; however, the supplied material contains no verifiable real output, and external build, cloud-session, and network costs prevent proving that only light review is needed, so the static ceiling applies.
The source identifies Expo documentation, eas CLI, and agent-device CLI as authorities and gives version and validation-window claims; however, there is no committed test suite, CI coverage, or independent execution record, leaving key claims primarily author-reported. The static-review ceiling therefore limits this to5 points.
- EAS Simulator is paid and experimental; obtain explicit confirmation of cost and duration before startup and stop sessions on every exit path.
- EXPO_TOKEN, daemon tokens, ARGENT_AUTH_TOKEN, public preview URLs, and artifact URLs can create credential or access-control risks; keep them out of repositories and uncontrolled logs.
- The commands use npx --yes with latest and depend on remote CLIs, EAS cloud infrastructure, and tunneling; version drift, network failure, or quota limits may break the workflow.
- The files do not establish reliable mainland-China reachability for Expo/EAS, tunnel services, or artifact downloads.
What does this skill do, and when should you use it?
This skill teaches an AI agent to use EAS Simulator to start remote iOS and Android simulator sessions and control apps through the CLI or agent-device. It is aimed at Linux, CI, cloud sandboxes, and macOS users who need a remote session or an iOS version they do not have locally. The workflow covers session startup, app installation, interaction, screenshots, browser preview, and cleanup. EAS Simulator is a paid service, and its simulator commands are experimental.
It checks whether a local simulator is available, then guides the agent through npx --yes eas-cli@latest simulator:start, simulator:get, simulator:exec, and simulator:stop. Through agent-device, it installs apps, opens bundle IDs or deep links, reads the accessibility tree, presses references or selectors, fills fields, and captures screenshots. It also selects among a local Release build, an EAS simulator build, and a development build connected to Metro, while managing .env.eas-simulator and session cleanup.
- A Linux developer needs to run an iOS Expo app without access to a local iOS simulator.
- A CI or cloud agent must launch a remote simulator and verify the app interface.
- A developer wants an AI agent to click through an app, enter text, and produce screenshots.
- A team wants to share a live iOS simulator preview through a browser URL.
- An Expo developer needs a development build with Fast Refresh for an agentic edit-and-see loop.
What are this skill's strengths and limitations?
- Designed for Linux, CI, and cloud environments without local simulators.
- Supports CLI control, AI-agent interaction, and an iOS browser preview.
- Clearly distinguishes static builds from development builds for live-refresh workflows.
- Includes practical checks for readiness, billing, orphaned sessions, and dotenv cleanup.
- Licensed under MIT.
- EAS Simulator is paid; remote sessions consume plan compute allowance and add latency.
- The simulator:* commands are experimental and hidden, so syntax may change.
- The documented browser preview is available for iOS, not Android.
- It requires network access, EAS authentication, Node.js-based CLIs, and filesystem writes.
- No test suite is provided in the source, and the source does not establish that every platform and command combination is tested.
How do you install this skill?
The README documents installation of the full Expo skills collection: npx skills@latest add expo/skills --skill '*'. For Claude Code, use claude plugin install expo@claude-plugins-official. For Codex, use codex plugin add expo@openai-curated. The provided source does not document a standalone installation command for only eas-simulator.
How do you use this skill?
From an Expo project directory, prompt the agent: "Run my Expo app on a cloud iOS simulator and take a screenshot." On non-macOS hosts or machines without a local simulator, proceed with EAS Simulator; on macOS with a local simulator, first establish whether the user wants cloud or local execution. A typical flow starts with npx --yes eas-cli@latest simulator:start --platform ios --type agent-device --non-interactive, uses npx --yes eas-cli@latest simulator:exec npx agent-device@latest ... to install and drive the app, and ends with npx --yes eas-cli@latest simulator:stop. The source says to read references/run-your-app.md before selecting a build mode, but that file's contents are not provided here.
How does this skill compare with similar options?
Compared with expo run:ios, Xcode, or Android Studio local simulators, this skill targets cloud and remote-device workflows. Compared with EAS Build, its core purpose is running and controlling simulator sessions rather than building or distributing apps.