# Mountspec

> Behaviour-first Vitest unit tests for one Vue 3 component, composable or Pinia store - written for you, or your
> existing tests reviewed - after the agent skill @github/unit-test-vue-pinia. https://mountspec.skillsafe.ai/
> Independent tool: not affiliated with or endorsed by the Vue, Pinia or Vitest projects.

## What it does

- Free, in the browser, no account: reads the pasted unit's public surface (defineProps, defineEmits, defineModel,
  slots, defineStore state / getters / actions, a composable's return values, the stores it uses), recommends the
  least powerful Pinia setup and a mount style, lints a pasted Vitest file, and exports a spec skeleton.
- Pinia strategies: testing_pinia (createTestingPinia({ createSpy: vi.fn }) while mounting - the default for a
  component that uses a store), testing_pinia_real_actions (stubActions: false, only when the real action must run),
  real_pinia (setActivePinia(createPinia()) for pure store tests), active_testing_pinia
  (setActivePinia(createTestingPinia(...)) for stores or composables with dependent stores), none.
- Lint rules: wrapper_vm, snapshot, async_update (unawaited trigger / setValue / setProps / setChecked), determinism
  (Date.now, Math.random without fake timers), no_assertion, mock_reset, one_behaviour, brittle_selector,
  implementation_detail (setData). Accepted minimal setups that are never flagged: createTestingPinia({}),
  createTestingPinia({ initialState }) or ({ stubActions }) without createSpy when no spy is inspected, and
  setActivePinia(createTestingPinia(...)) in store or composable tests.
- Paid lane "write" (gpt-terra): the behaviour boundary, Pinia strategy and mount, three to ten cases named for the
  behaviour they prove, a complete Vitest file with one it() per case, and the gaps left open.
- Paid lane "review" (gpt-terra): keep / tighten / rewrite, severity-ranked findings with rule, line, problem and fix,
  the accepted setups, missing coverage (handed to "write" as its focus) and brittleness risks.
- Every reply is checked against the page's reading: surface ids must exist in the source, the written file is re-linted,
  its it() blocks must match the cases, and its Pinia setup must match the declared strategy.

## API

- Tutorial with samples in eight languages: https://mountspec.skillsafe.ai/api.html
- Base URL https://api.skillsafe.ai/v1/app-api - /guest, /me, /estimate (free), /run, /run-stream.
- Input (all strings): task (write | review), title, source, tests (review), focus, facts (JSON string built by
  VueKit.buildInput in https://mountspec.skillsafe.ai/vuekit.js), retry_note.

## Sources

- Agent skill: https://skillsafe.ai/skill/@github/unit-test-vue-pinia (github.com/github/awesome-copilot,
  skills/unit-test-vue-pinia, MIT), with its references/pinia-patterns.md.
- Pinia testing cookbook: https://pinia.vuejs.org/cookbook/testing.html
- Vue Test Utils guide: https://test-utils.vuejs.org/guide/