| name | rn-write-tests |
| description | Use to write tests for an Expo + RN app: Jest + React Native Testing Library for unit/integration (components, hooks, queries, mutations) and Maestro for end-to-end flows (sign-in, navigation, forms). Sets up the testing stack on first call (jest-expo preset, RNTL, jest config) and writes a focused test next to the source file. Triggers on: "write tests for X", "add e2e test", "mock expo-notifications", "test this hook". Not for: choosing what to test (the user decides), running the existing test suite (just `npm test`), or testing pure logic outside an Expo project. |
rn-write-tests — Jest + RNTL + Maestro testing for Expo + RN
Contract
See references/contracts.md (vendored from dev-flow). Key facts:
- Reads
<project-root>/.workflow/meta.json#stack.framework — must be "expo-rn".
- Adds dev dependencies + Jest config on first call (idempotent: detects existing setup).
- Writes test files under
<project-root>/__tests__/ (Jest convention) for unit/integration.
- Writes Maestro YAML flows under
<project-root>/.maestro/ for e2e.
- Does NOT modify
meta.json#phase.
When this skill applies
- User asks to write a test for a specific source file or flow.
- Orchestrator does NOT route here automatically — test writing is on-demand.
Knowledge dependencies (read these first)
rn-fundamentals/SKILL.md — TS strict, modern primitives the tests will see.
rn-data-fetching/references/patterns.md — query/mutation patterns the tests must exercise.
Workflow
Step 1 — Verify preconditions
Read .workflow/meta.json. Abort if stack.framework != "expo-rn".
Step 2 — Detect / install the test stack
Check if jest, jest-expo, @testing-library/react-native are in package.json. If not:
npx expo install --dev \
jest jest-expo \
@testing-library/react-native \
@testing-library/jest-native \
@types/jest -- --legacy-peer-deps
(See references/jest-setup.md for the full config.)
For Maestro: detect .maestro/ directory. If missing AND user wants e2e, see references/maestro-setup.md (manual install, not npm).
Step 3 — Choose what to test
Ask the user (one round-trip) what to test:
- Component (renders, props, interactions) → RNTL.
- Hook (
useQuery, useMutation, custom hook) → RNTL renderHook.
- Pure function (utility, helper) → plain Jest.
- e2e flow (sign-in, navigation, form submit) → Maestro.
If the file under test is lib/api.ts, write a Jest test of the function with fetch mocked.
If the file under test is app/(auth)/sign-in.tsx, write an RNTL test of the screen + a Maestro flow of the user journey.
Step 4 — Write the test
Use the canonical patterns from references/rntl-patterns.md (or references/jest-setup.md for utilities, or references/maestro-setup.md for e2e).
Test files:
- Unit/integration:
__tests__/<mirror-source-path>.test.tsx (or .test.ts for non-JSX).
- e2e:
.maestro/<flow-name>.yaml.
Step 5 — Run the test
cd <project-root> && npm test -- --runInBand --bail
(For Maestro: maestro test .maestro/<flow-name>.yaml.)
Test MUST pass. If it doesn't:
- Read the failure. Is it a missing mock? An assumption about the test API?
- Fix the test (NOT the source — that's a separate task).
- Re-run.
Step 6 — Commit
git add __tests__/<path>.test.tsx
git commit -m "test(<area>): cover <what>"
Common anti-patterns (NEVER do)
- ❌ Detox — heavy, Expo Go unfriendly, requires native build. Use Maestro.
- ❌ Enzyme — abandoned. RNTL only.
- ❌ Snapshot for every component — only for stable design-system primitives.
- ❌
jest.fn() without typing the return → tests pass on wrong shape.
- ❌ Testing implementation details (state names, internal function calls) → brittle. Test user-facing behavior.
- ❌ Testing TanStack Query result shape directly → mock the queryFn instead and assert UI.
- ❌ Skipping cleanup → tests leak state. Always use RNTL's auto-cleanup (default).
- ❌ Sleeping
await new Promise(r => setTimeout(r, 1000)) → use waitFor with explicit assertions.
Updating meta.json (recommended pattern)
When this skill modifies state (artifact written, phase advanced, history appended), use the canonical script when available:
python3 .../dev-flow/scripts/update_meta.py <project-root> record-artifact \
--path <relative-path> --produced-by '<this-skill-name>' [--derived-from <p1> <p2> ...]
python3 .../dev-flow/scripts/update_meta.py <project-root> set-phase <new_phase>
python3 .../dev-flow/scripts/update_meta.py <project-root> append-history \
--skill '<this-skill-name>' --inputs '{...}' --outputs '{...}' --phase-after <new_phase>
The script enforces phase monotonicity, normalizes legacy kebab-case aliases (e.g. module-added → module_added), and writes the canonical sha256 + timestamp into meta.json#artifacts. Fall back to direct JSON editing only if the script is not on PATH (and warn the user).
Sources