name: e2e-frontend-validation description: E2E validation workflow for frontend changes in playground packages using Playwright MCP model: claude-opus-4-5

E2E Validation for Frontend Modifications

Prerequisites

Requires Playwright MCP server. If the browser_navigate tool is unavailable, instruct the user to add it:

claude mcp add playwright -- npx @playwright/mcp@latest

Validation Steps

After completing frontend changes:

1. Build the CLI

pnpm build:cli

2. Start the dev server

cd packages/playground/e2e/kitchen-sink && pnpm dev

3. Verify server is running

   • URL: http://localhost:4111
   • Wait for the server to be ready before proceeding

4. Identify impacted routes

   • Routes are defined in packages/playground/src/App.tsx
   • Browse them ALL to verify behavior

5. Test with Playwright MCP

   • Use browser_navigate to visit each impacted route
   • Visually verify the changes render correctly
   • Test any interactive elements modified
   • Use browser_screenshot to capture results for the user

6. Write E2E Tests

   Write Playwright tests covering the validated behavior. Follow the conventions below.

   File Placement

   Map validated routes to test file paths under packages/playground/e2e/tests/:

   Route pattern	Test file
   /agents	agents/page.spec.ts
   /agents/:id/...	agents/$agentId/page.spec.ts
   /agents/:id/chat/:chatId	agents/$agentId/stream.spec.ts (if streaming)
   /tools	tools/page.spec.ts
   /tools/:id	tools/$toolId/page.spec.ts
   /workflows	workflows/page.spec.ts
   /workflows/:id	workflows/$workflowId/page.spec.ts
   /mcps	mcps/page.spec.ts
   /mcps/:id/tools/:toolId	mcps/$serverId/tools/$toolId/page.spec.ts

   • Layout/navigation tests → page.spec.ts
   • Streaming/chat tests → stream.spec.ts

   New File vs Extend Existing

   • If the spec file already exists, add new test() blocks to it.
   • If it doesn't exist, create a new file with the full boilerplate (imports, afterEach hook, etc.).

   Fixture Decision Tree

   • AI streaming responses (text-delta, tool calls, workflow execution) → use selectFixture() + nanoid() for deterministic, isolated tests
   • Static UI, navigation, forms, deterministic tool execution → no fixture needed

   Test Conventions

   // Imports
   import { test, expect } from '@playwright/test';
   import { resetStorage } from '../__utils__/reset-storage';
   // Only for streaming tests:
   import { selectFixture } from '../__utils__/select-fixture';
   import { nanoid } from 'nanoid';
   

   • Always call resetStorage() in test.afterEach
   • For streaming tests, create a fresh browser context in test.beforeEach:

     let page: Page;
     test.beforeEach(async ({ browser }) => {
       const context = await browser.newContext();
       page = await context.newPage();
     });
     

   • Locator priority: getByTestId > getByRole > getByLabel > locator('text=...')
   • Use { timeout: 20000 } for async/streaming content assertions
   • Use nanoid() for unique chat session IDs in URLs
   • Verify memory persistence with page.reload() where applicable
   • Use toMatchAriaSnapshot() for stable layout assertions

   Kitchen-Sink Extension

   When the feature under test requires new fixtures or entities:

   • New fixtures: Add to e2e/kitchen-sink/fixtures/, create a <name>.fixture.ts file, and register it in fixtures/index.ts
   • Update types: Add the fixture name to the Fixtures union in e2e/kitchen-sink/types.ts and e2e/tests/__utils__/select-fixture.ts
   • New tools/agents/workflows: Add to the corresponding files in e2e/kitchen-sink/src/mastra/

7. Verify Tests Pass

   cd packages/playground && pnpm test:e2e
   

   If tests fail, fix them and re-run until green.

Quick Reference