test-looplisted

# Test Automation Loop > The core insight: if the agent is going in circles, it needs a test — not more prompts. Write the test. Let the agent iterate against it. Walk away. ## The testing.md Pattern Every project should have a `testing.md` at the root (or in relevant submodule folders) containing EVERYTHING the agent needs to run tests autonomously: ### Template: ```markdown # Testing Guide ## Environment Setup - Required env vars: [list with descriptions] - Database: [commands to start/seed test DB] - Services: [docker-compose, external APIs] - Test user: [how to create/seed test data] ## Running Tests ### Unit Tests Command: `npm test` Location: `tests/unit/` ### Integration Tests Command: `npm run test:integration` Env vars needed: [list] What they test: [scope description] ### E2E Tests (Playwright) Command: `npx playwright test` Setup: `npx playwright install chromium` Base URL: [how it's configured] Auth: [how test auth works] Location: `tests/e2e/` ## Debugging Failed Tests - Single test: `npm test -- -t "test name"` - Headed browser: `npx playwright test --headed` - Traces: `npx playwright show-trace test-results/*/trace.zip` - Verbose: `npm test -- --verbose` ``` **Create this file FIRST** if it doesn't exist. The agent cannot iterate autonomously without it. ## The Autonomous Loop ``` 1. Read testing.md 2. Set up the test environment (DB, env vars, services) 3. Ensure screenshot capture is enabled: - Verify playwright.config.ts has screenshot: 'on'