coatcheck 0.2.1

Contracts and a Playwright accessibility runner for Topcoat component workbenches.
Documentation
# Runner

`coatcheck test` is intentionally manifest-first:

1. optionally start `host.command`;
2. wait for `host.health`;
3. fetch and validate `host.manifest`;
4. generate one explicit Playwright test per story/scene under
   `.coatcheck/generated`;
5. run generated and project-local specs with Bun;
6. write semantic results and review artifacts.

The generated gate checks successful navigation, the configured root and
ready markers, absence of the host error marker, uncaught page errors, browser
console errors, and configured Axe impacts.

Every manifest case writes:

- `review.png` for human review;
- `dom.html`;
- `axe.json`;
- `console.json`.

The PNG is never compared with a pixel baseline. Pass/fail comes from semantic
and accessibility checks. Playwright traces are retained for failures.

Custom Playwright specs are ordinary project files matching
`test.story_globs`. They use the project-local `@playwright/test` installed by
`coatcheck init --package-manager bun`.