test-automationlisted

# Test Automation ## Environment (this machine) - Python venv: `~/qa-agent/.venv/bin/python` — has pytest, playwright, requests, locust. - Dispatcher: `qa` CLI (`qa test`, `qa smoke`, `qa regression`, `qa security`, `qa report`). - Chrome MCP tools are available for live-browser automation/inspection alongside scripted Playwright. ## What to automate (and what not to) Automate: - Regression-prone flows run repeatedly (login, checkout, core CRUD). - Anything that gates a release (smoke suite). - API contract checks — cheap to write, fast to run, high signal. Don't automate (or automate last): - One-off exploratory checks. - Highly visual/subjective UI polish — use screenshots + manual review instead. - Flows still actively changing week to week (automation will just be a maintenance drag). ## Playwright test structure - **Page Object Model**: one class per page/component encapsulating locators + actions. Tests read like user stories, not selector soup. - **Locators**: prefer role/text/testid locators over CSS/XPath — they survive markup churn. Add `data-testid` to the app if none exist and you own the code. - **Waits**: never use hard `sleep()`. Use Playwright's auto-waiting or explicit `wait_for_selector` / `wait_for_load_state("networkidle")`. - **Isolation**: each test should set up its own data (via API/fixtures) and not depend on execution order or another test's leftover state. - **Assertions**: assert on the outcome (DB state, API response, final DOM state), not int