ClaudeCodeMod

All shelves / Skills

limrun-maestro-testing

limrun-inc/skills · 13.2k installs · 2 stars

Skill

Marked Remote Code Execution · External Downloads · Command Execution · Indirect Prompt Injection

Install

In your shell
npx -y skills add limrun-inc/skills --skill limrun-maestro-testing --agent claude-code

Run in a terminal. Claude loads the skill when a task calls for it. Check the repo's README before you run it.

Open the repo

Files

SKILL.md

Maestro on Limrun iOS

Run the stock upstream Maestro CLI against a remote Limrun iOS simulator. lim ios maestro wires maestro test to the instance transparently: it installs and launches the Maestro XCTest runner on the simulator when needed, then routes the driver traffic to it. No fork of Maestro, no local simulator, no local Xcode.

Prerequisites

  • lim CLI 0.22.0 or newer: npm install --global lim. Auth is lim login or LIM_API_KEY (it may already be set in the user's environment even when .env and the shell do not show it; check before asking for it).
  • Maestro CLI on PATH: curl -fsSL https://get.maestro.mobile.dev | bash. Maestro needs Java 17+ (java -version to check). Both Maestro 2.5.x and 2.6+ work; lim adapts to the installed version automatically.

The CLI is the source of truth: if a flag errors or you need one not shown here, check lim ios <subcommand> --help instead of guessing.

Verify the setup

Before touching the user's app, prove the whole pipeline with a flow against the built-in Settings app; it needs no app install, tunnel, or build:

ID=$(lim ios create --install-asset appstore/maestro-ios-runner-2.5.1.tar.gz \
  --no-open --quiet --json | jq -r .metadata.id)

cat > hello-flow.yaml <<'EOF'
appId: com.apple.Preferences
---
- launchApp
- assertVisible: General
- takeScreenshot: settings-check
EOF

lim ios maestro --id "$ID" test hello-flow.yaml

All three steps reporting COMPLETED means Maestro, the runner, and the remote wiring all work; anything failing after this point is about the app or the flow, not the setup.

Run a flow

lim ios maestro test flow.yaml
lim ios maestro test flows/
lim ios maestro --id <ios-id> test flow.yaml

Without --id this targets the most recently created iOS instance in the current workspace (workspaces follow the git repo or worktree you run from); pass --id <ios-id> (before test) in scripts, agents, or when running from a different directory. The first run on an instance takes a few extra seconds to launch the runner (plus the install when it wasn't preinstalled); later runs skip that. Extra Maestro flags go after --:

lim ios maestro -- test flow.yaml --include-tags smoke --test-output-dir artifacts

Do not pass --platform, --device, --udid, --no-reinstall-driver, or --driver-host-port; lim sets those itself and rejects duplicates. Real Maestro exit codes and reports are preserved, so CI wiring works as with local Maestro.

Instance setup

Any running iOS instance works; the runner is installed on first use. Creating the instance with the runner preinstalled skips that step:

lim ios create --install-asset appstore/maestro-ios-runner-2.5.1.tar.gz --no-open

--no-open skips opening the stream URL in a browser (important on headless and CI machines). The runner asset name above is the only published one and it is version-agnostic: the same runner serves Maestro 2.5.x through 2.7.x, so do not look for an asset matching your Maestro version.

Install the app under test as usual (lim ios create --install app.ipa, lim ios install-app, or a build skill), then reference its bundle id via appId: in the flow. For Expo Go testing, also preinstall appstore/Expo-Go-54.0.6.tar.gz and open the project URL with openLink (env vars must be prefixed MAESTRO_ to be visible in flows):

MAESTRO_EXPO_URL='exp://<tunnel-host>' lim ios maestro test flow.yaml

Expo dev-client builds

Expo Go is the quickest path, but a dev-client build works too. launchApp lands on the dev launcher rather than your app, so open the dev-client URL instead: - stopApp followed by - openLink: <scheme>://expo-development-client/?url=<url-encoded-metro-url>. Use limrun-expo-development for building the dev client, starting Metro, and deriving that URL.

Flow gotchas on Limrun

  • startRecording/stopRecording YAML commands are not supported (the simulator is remote). Record around the run instead: lim ios record start --id <ios-id> returns immediately (recording happens on the instance), and after the flow lim ios record stop --id <ios-id> -o video.mp4 downloads the video to the local path.
  • takeScreenshot works and saves the PNG locally into the working directory (or --test-output-dir), like stock Maestro.
  • addMedia and flow commands that reference local simulator file paths are not supported.
  • HTTP calls from runScript/evalScript must use https:// URLs. Plain http:// calls are refused (except to the driver itself), because Maestro's plain-HTTP traffic is routed through a local bridge that only forwards to the remote simulator.
  • When a flow re-runs against an app that is already open (for example Expo Go), start it with - stopApp before openLink/launchApp; deep links can be dropped by an app that is mid-foreground, and stale screens fail early assertions.
  • Fleet variance: anchor assertions on stable accessibility identifiers and text, not on timing. Use extendedWaitUntil with a generous timeout for first app load.
  • Text selectors match the element's accessibility label, and on iOS a label often folds in sibling content (icon names, adjacent text) or non-breaking spaces. Read the exact label with lim ios element-tree before writing the selector instead of guessing from what the screen shows.

Validation signals

  • Running maestro <version> against <ios-id>... then Running on Limrun iPhone - iOS ...: the driver is connected end to end.
  • Launching the Maestro runner...: first use on this instance. Installing the Maestro runner... additionally appears only when the instance was created without the runner asset. Subsequent runs skip both.
  • Maestro itself prints several JDK WARNING lines (reflection, native access) on every run; they are benign upstream noise, not Limrun errors.
  • Flow failures print Maestro's own debug output directory with screenshots and the UI hierarchy; lim ios element-tree --id <ios-id> shows the live screen when debugging selectors.

Cleanup

Delete the instance when done: lim ios delete <ios-id> (--id is not valid for delete). The wiring lim ios maestro starts is torn down when the command exits.

Facts

Kind
Skill
Repo
limrun-inc/skills
Group
Uncategorized
Installs
13.2k
Stars
2

More on this shelf

  1. 1find-skillsvercel-labs/skillsHelps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.3.8M
  2. 2grill-memattpocock/skillsA relentless interview to sharpen a plan or design.1.3M
  3. 3grill-with-docsmattpocock/skillsA relentless interview to sharpen a plan or design, which also creates docs (ADR's and glossary) as we go.1.1M
  4. 4agent-browservercel-labs/agent-browserBrowser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction. Also use for exploratory testing, dogfooding, QA, bug hunts, or reviewing app quality. Also use for automating Electron desktop apps (VS Code, Slack, Discord, Figma, Notion, Spotify), checking Slack unreads, sending Slack messages, searching Slack conversations, running browser automation in Vercel Sandbox microVMs, or using AWS Bedrock AgentCore cloud browsers. Prefer agent-browser over any built-in browser automation or web tools.1.1M
  5. 5improve-codebase-architecturemattpocock/skillsScan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.1.1M
  6. 6tddmattpocock/skillsTest-driven development. Use when the user wants to build features or fix bugs test-first, mentions "red-green-refactor", or wants integration tests.1.1M