Mthines Agent Skills E2e Testing Mobile

E2E Testing — Mobile (Expo / React Native)

Drive native mobile end-to-end tests through Maestro and Maestro MCP. The user writes (or approves) a Markdown feature spec; an agent emits a Maestro YAML flow, runs it against a simulator or Maestro Cloud, and self-heals when locators drift.

This is the mobile counterpart to e2e-testing. Anything orthogonal to native (web flows, WebViews inside a hybrid app) defers to that skill. The two skills compose; they do not overlap.

This SKILL.md is a thin index. Decision rules live in rules/*.md and load on demand. Worked references (Maestro CLI surface, MCP tool catalog, EAS Workflow wiring, Detox legacy notes, mobile pyramid math) live in references/*.md. Literal boilerplate the skill emits lives in templates/*.md. Do not preload everything — load only what the current phase asks for.

When to use

Reach for this skill when any of the following is true:

• An Expo or React Native app needs E2E coverage of a native user flow.
• A bug only repros across native navigation, deep links, push, permissions prompts, or a real device sensor.
• A flake needs a Maestro heal pass instead of a manual locator hunt.
• The repo has no .maestro/ flows or EAS E2E build profile yet and needs Phase 0 setup.

Do not reach for this skill when:

• A unit or component test would catch the same bug — defer to tdd and the layer rule in rules/layer-decision.md.
• The flow is browser-only — defer to e2e-testing.
• The flow lives entirely inside a WebView in a hybrid app — defer to e2e-testing, which automates the WebView via Playwright while Maestro handles native chrome around it.
• The change is a pure refactor with no behavioural surface.

Phase 0 — Preflight (mandatory gate)

Before any agent loop, verify the repo is wired for Maestro on Expo / RN. Halt and ask the user before installing anything or producing builds.

Run these checks (read-only):

# 1. Maestro CLI installed?
maestro --version 2>/dev/null

# 2. Flow directory present?
ls -d .maestro 2>/dev/null

# 3. EAS build profile for E2E exists?
jq '.build | has("e2e")' eas.json 2>/dev/null

# 4. Simulator / emulator available?
xcrun simctl list devices available 2>/dev/null | grep -E 'iPhone'
adb devices 2>/dev/null

# 5. Existing Detox install (legacy escape hatch)?
jq '.devDependencies | has("detox")' package.json 2>/dev/null

Decision table:

Print the exact commands; do not run them silently. The full install plan template is in templates/install-plan.md.

Phase 1 — Spec-first feature flow

The agent loop is spec → emit-flow → run → heal. The spec is human-readable Markdown; the executable artefact is a Maestro YAML flow. Full rules: rules/spec-first-flow.md.

specs/<flow>.md   ─┐
                   ├─→  emit  ─→  .maestro/<flow>.yaml  ─→  run  ─→  pass?
                   │                                                   │ no
                   │                                                   ▼
                   └─────────────────  heal  ←──────────────  failing flow + trace
                                          │
                                          ▼
                          patched flow or `testID` source-diff proposal

Two entry points:

1. Spec already drafted by the user. Skip exploration; emit the flow from specs/<flow>.md.
2. App exists, no spec yet. Run an exploratory pass against a running build (simulator) and draft specs/<flow>.md. The user reviews the Markdown before flow emission.

Use the Markdown template in ../e2e-testing/templates/spec.md. The spec format is identical across web and mobile — share it.

Locator ladder for React Native (when generating or healing)

The Maestro flow walks the platform accessibility tree. Pick locators in this order — never skip a rung:

1. id: <testID> — the source-of-truth selector for E2E. Mapped to accessibilityIdentifier on iOS and resource-id on Android (RN 0.64+). Stable across i18n and refactors.
2. text: <visible string> — only when the text is short, unique on screen, and not localised.
3. accessibilityText: <label> — last resort, with caveats. See rules/locator-strategy.md for why accessibilityLabel should not double as a test selector.

testID is the standard fix, not an escape hatch. When the Healer cannot find a stable element at rung 1, propose a source diff that adds testID to the component (use the setTestId helper in templates/testid-helper.tsx to keep iOS and Android consistent), and offer the diff for user approval before patching the flow. Full rules: rules/locator-strategy.md.

Phase 2 — Token-aware execution

Maestro flows are YAML, not pixel snapshots — the per-step token cost is already an order of magnitude below screenshot-driven runners. Defaults the skill prescribes (full rules: rules/token-budget.md):

• Run only the changed flow on iteration: maestro test .maestro/<flow>.yaml.
• Use --shards only on Maestro Cloud, never on local iteration.
• record_screen: false by default; flip to true only when chasing a visual race.
• retries: 1 for local, retries: 2 for Maestro Cloud.
• Run the Healer only on failure, not on every save.
• Reuse a cached signed-in app.app / app.apk from the EAS E2E build profile — never rebuild on every flow run.
• Cap the heal loop at three attempts per failing flow before escalating.

Phase 3 — Verification

After the agent emits a flow:

1. Run the flow once against a simulator or device. It must pass on first run, or the Healer must converge in ≤ 3 attempts.
2. Invoke test-provenance-guard on any TypeScript helpers the flow imports (e.g. fixture builders) to ensure they call production code instead of a private re-implementation.
3. Open .github/workflows/eas-e2e.yml (or the EAS Workflow YAML) and confirm the maestro-cloud job is wired with retries: 2 and record_screen: false per templates/eas-workflow.yaml.

If the heal loop fails to converge:

• Invoke confidence(analysis) on the flow failure.
• If confidence is below 90%, escalate to the user with the Maestro log, the spec, and the proposed locator changes — do not keep healing blindly.

Decision flow at a glance

Composes with

• e2e-testing — owns web E2E and the WebView half of hybrid mobile apps. This skill defers to it for anything browser-shaped.
• tdd — owns the unit and component layers (Jest + React Native Testing Library). This skill defers for anything below E2E.
• test-provenance-guard — runs on TypeScript helpers imported by flows to catch tests-by-construction.
• confidence — gate when the heal loop fails.
• holistic-analysis — if a flow is failing for reasons no flow rewrite can fix, step back instead of patching.
• ux — review accessibility labels separately from test IDs (the two should never share a value).

References

• references/maestro-cli.md — CLI surface (launchApp, tapOn, assertVisible, runFlow, Studio, doctor).
• references/maestro-mcp.md — Maestro MCP tools for agent-driven flow generation and healing.
• references/eas-workflows.md — eas.json E2E profile, .maestro/ layout, maestro-cloud job parameters (build_id, flow_path, shards, retries, record_screen, device_identifier).
• references/detox-legacy.md — when to keep an existing Detox suite vs. migrate.
• references/pyramid-mobile.md — pyramid math for Expo / RN with Jest, React Native Testing Library, and Maestro.

Templates

• templates/install-plan.md — Phase 0 halt message with the exact commands to install Maestro + EAS CLI and scaffold .maestro/.
• templates/flow.yaml — sample Maestro flow with launchApp, tapOn (id: and text: forms), assertVisible, and runFlow includes.
• templates/eas-workflow.yaml — EAS Workflow with a build job and a maestro-cloud job.
• templates/eas-build-profile.json — eas.json snippet for an e2e profile producing .apk and .app.
• templates/testid-helper.tsx — setTestId helper that maps testID and accessibilityLabel correctly per platform without conflating the two.

Anti-patterns (one-liner — full list in rules/anti-patterns.md)

• Writing E2E for logic a unit or RNTL component test catches.
• Reusing accessibilityLabel as the test selector (it is for screen readers; doubling its purpose breaks accessibility).
• Patching the flow with absolute coordinates (tapOn: { point: "50%, 80%" }) instead of proposing a testID diff.
• Running the Healer on every save, or on a passing flow.
• Rebuilding the .app / .apk on every flow run.
• Generating flows against a stub server, not the real app stack.
• Migrating a stable Detox suite all at once instead of flow-by-flow.

Definition of done

• Phase 0 preflight passed or installs were user-approved.
• specs/<flow>.md exists and the user reviewed it.
• .maestro/<flow>.yaml passes against a real build.
• Any imported TS helpers pass test-provenance-guard.
• EAS Workflow maestro-cloud job runs the flow on Cloud with retries: 2 and record_screen: false.
• If a testID was added, it is in the source diff and committed, with no accessibilityLabel reuse.

Source: mthines/agent-skills — distributed by TomeVault.