automation-apilisted
Install: claude install-skill YevheniiaQA/qa-agent-toolkit
# Automation API (TypeScript + Playwright)
Generate tests a senior engineer would approve in review — typed, isolated, assertive,
traceable — not code that merely runs.
## Project shape
One Playwright project for every API test in the product, `03-tests/api/`, with typed clients,
Zod response schemas and fixtures under `src/`, and specs under `specs/<feature>/` — by feature,
never by ticket; a new ticket adds spec files, it never creates a second project. Read
`references/project-and-config.md` before creating or extending the project: the layout, the
config baseline, the base-URL trap, and what to read in a target repository first.
## Generation pipeline
**Phase 1 — setup.** Load the scenario JSON, the relevant API files from the
global context, and resolved test data. Load only the service contracts this
ticket touches; loading the whole context degrades output quality.
**Phase 2 — generate.** One `test()` per scenario. The test title carries the
scenario ID so a CI failure maps to a requirement without a lookup.
**Phase 3 — debug.** Hand off to `test-debugger`. Do not declare completion on
code that has never executed.
**Phase 4 — report.** Emit `DEBUG_REPORT.md` with pass/skip/expected-failure
counts and every classification decision.
## Code standard
```ts
import { test, expect } from '../src/fixtures';
import { ConnectResponse } from '../src/schemas/integrations';
test.describe('US-001 Connect a Confluence workspace', () => {
test('TC-283-API-001 connects w