0xerwin1/nixos/setup-testing

When to Use

Run this skill when:

• The user asks to set up the testing environment (setup testing, install testing dependencies, preparar el entorno de testing).
• The testing pipeline (explore-testing, plan-testing, run-testing) detects missing prerequisites and proposes setup before continuing.
• A fresh clone or a new machine needs to become test-ready.

Do NOT run this skill for: test execution, test planning, or fixing test failures unrelated to missing tools.

Operating Principles

• Discover first. Identify OS, shell, package managers, existing tools, connected devices, and repo hints before proposing any action.
• Stay system-agnostic. Support Linux, macOS, Windows, and WSL. Do NOT assume Homebrew, apt, winget, npm, Android Studio, Xcode, or any specific package manager is available.
• Ask before heavy actions. Always get confirmation before installs, SDK downloads, accepting Android licenses, creating AVDs/simulators, using cloud devices, or writing repo files.
• Keep source safe. Do NOT edit application source. Only write testing support files when the user explicitly approved that scope.

Detection Phase

Before proposing any action, detect current state inline:

1. Detect platform and shell context:

• Linux / macOS / Windows / WSL
• available shells / terminals
• available package managers or installers (if any)

1. Check repo hints:

• TESTING_SETUP.md
• .maestro/ directory or existing *.yaml flows
• package.json / lockfiles / existing test scripts
• .mcp.json or other MCP configuration files if present

1. Check Playwright readiness:

• @playwright/test in package.json devDependencies when applicable
• npx playwright --version
• npx playwright install --dry-run

1. Check Maestro readiness:

• maestro --version
• whether the runtime exposes Maestro MCP helpers (list_devices, run, etc.)
• existing .maestro/**/*.yaml flows

1. Check browser / web-surface readiness when Playwright or Maestro web validation matters:

• command -v google-chrome || command -v chromium || command -v chrome
• Claude CLI / extension bridge availability if the user wants the chrome-extension engine (Claude Code only, when the Claude Chrome extension is available)
• any repo-documented browser launch command or approved web target for Maestro (the web-surface path on other clients)

1. Check device tooling when Maestro is relevant:

• adb devices
• emulator -list-avds when Android emulator tooling exists
• xcrun simctl list devices on macOS when Xcode tools exist
• user-configured cloud device provider only if already present in repo docs or environment

1. Check app targets when mobile is relevant:

• appId, bundle ID, app path, or launch command from TESTING_SETUP.md
• obvious repo hints (e.g. AndroidManifest / Info.plist / app config) without inventing values

If a command is unavailable on the current OS, record that as a discovery, not as a failure.

Summary and Confirmation

After detection, report what is present, what is missing, and what setup options are possible on this machine. Use plain language and group the result by engine / surface:

• Playwright (package + browser binaries)
• Maestro (CLI + MCP helpers + flows)
• Browser / web-surface path (Claude Code: chrome-extension engine when available; other clients: repo-documented browser launch or approved web target)
• Devices / simulators / emulators
• App targets / credentials / env vars

Then ask what the user wants to do next. Example:

"Para el pipeline de testing veo esto:

- Playwright: instalado, pero faltan binarios de WebKit

- Maestro: CLI no encontrado; tampoco veo helpers MCP activos

- Android: adb está, pero no hay ningún device conectado

- iOS: no aplica en esta máquina

- Browser / web target: no veo un navegador Chromium listo para la ruta visual

Puedo preparar Playwright, instalar Maestro, o dejarlo en best-effort. Antes de instalar nada pesado, aceptar licencias, crear emuladores, o escribir archivos del repo, te pregunto. ¿Cómo querés seguir?"

There is no default yes for installs or provisioning. Wait for the user's explicit approval.

Execution

After confirmation, do the smallest approved setup only.

• Playwright
• Install the package only if repo writes are approved.
• Install browser binaries only if the user approved the download.
• Maestro
• Install or upgrade the CLI using the user's approved toolchain.
• If Maestro MCP configuration is missing and the user approved repo/config writes, add or update it.
• Do NOT assume a package manager; if none is available, provide manual install steps instead of guessing.
• Devices / simulators / emulators
• Ask before downloading SDK components, accepting Android licenses, creating AVDs, booting emulators, creating iOS simulators, or using cloud devices.
• If the user says no, stop at discovery and report the remaining gap.
• Repo files / flows
• Ask before writing .maestro/**/*.yaml, .mcp.json, or other repo files.
• Prefer repo-provided setup scripts only when they exist, their actions match the approved scope, and the user approved running them.

After Completion

Report exactly what changed and what still needs manual action. Always call out:

• remaining credentials / env vars still missing
• whether Playwright browser binaries were installed
• whether Maestro CLI and MCP helpers are ready
• whether any device / simulator / emulator still needs the user to connect, start, or approve it
• whether the browser / web-surface path (chrome-extension on Claude Code, or any other browser/web target) still needs a manual start, login, or approval step

If the user chose not to install or provision something, say so explicitly rather than implying setup is complete.

Engram Save

After a successful run, save to engram:

mem_save(
  title: "Setup testing completed",
  type: "config",
  topic_key: "testing/{project}/setup-state",
  content: "What was installed, what was skipped, what requires manual action."
)