Webapp Testing

/supergraph:webapp-testing

Test and verify local web application UI behavior using Playwright.

Announce: "🌐 /supergraph:webapp-testing — testing web application..."

When to Use

• Verifying a frontend change works correctly in the browser
• Debugging UI behavior or unexpected rendering
• Capturing screenshots for visual verification
• Checking browser console logs for errors
• After /supergraph:frontend-design to confirm feature works end-to-end

Steps

0. Announce

"🌐 /supergraph:webapp-testing — testing [feature/page]..."

1. Understand what changed (graph context)

Before writing tests, use graph context to prioritize coverage: Use CBM_PROJECT: call detect_changes, search_graph for changed routes/pages, then trace_path for inbound/outbound call and data-flow paths. Run validated test-gaps. Empty results are unavailable; use Serena/filesystem evidence.

Serena (optional): For changed component symbols:

mcp__serena__find_referencing_symbols(symbol=<changed_component>)

If a Playwright test fails on a renamed component/route, find_referencing_symbols surfaces all import sites. Skip if Serena unavailable.

2. Choose approach

Is it static HTML?
├─ Yes → Read HTML file directly to identify selectors
│         Then write Playwright script using those selectors
│
└─ No (dynamic webapp) → Is server already running?
    ├─ No → Run: python scripts/with_server.py --help
    │        Then start server + write Playwright script
    │
    └─ Yes → Reconnaissance-then-action:
        1. Navigate and wait for networkidle
        2. Screenshot or inspect DOM
        3. Identify selectors from rendered state
        4. Execute actions with discovered selectors

MCP Playwright option: If mcp__plugin_playwright_playwright__* tools are available, prefer them over writing Python scripts — no script file needed:

mcp__plugin_playwright_playwright__browser_navigate(url="http://localhost:3000")
mcp__plugin_playwright_playwright__browser_snapshot()
mcp__plugin_playwright_playwright__browser_click(selector="...")
mcp__plugin_playwright_playwright__browser_take_screenshot()
mcp__plugin_playwright_playwright__browser_console_messages()

3. Recon before action (dynamic apps)

page.wait_for_load_state('networkidle')  # CRITICAL: wait before inspecting
page.screenshot(path='/tmp/inspect.png', full_page=True)
content = page.content()
page.locator('button').all()

Never inspect DOM before networkidle on dynamic apps — selectors may not exist yet.

4. Execute test actions

Use discovered selectors from step 3. Write targeted Playwright assertions.

5. Verify flows covered

After tests pass, re-run: Re-run search_graph then trace_path(project=CBM_PROJECT, mode="data_flow"). All impacted flows tested? Any gaps → add tests.

6. Report

✅ /supergraph:webapp-testing
- Flows tested: N/M | Console errors: [list/none]
- Screenshots: [paths] | Selectors verified: [list]
- Next: /supergraph:verify → /supergraph:review

Using with_server.py

Helper script scripts/with_server.py manages server lifecycle. Always run --help first — do NOT read the source unless absolutely necessary (large file, pollutes context window).

Single server:

python scripts/with_server.py --server "npm run dev" --port 5173 -- python your_test.py

Multiple servers:

python scripts/with_server.py \
  --server "cd backend && python server.py" --port 3000 \
  --server "cd frontend && npm run dev" --port 5173 \
  -- python your_test.py

Automation script template (server managed externally by with_server.py):

from playwright.sync_api import sync_playwright

with sync_playwright() as p:
    browser = p.chromium.launch(headless=True)
    page = browser.new_page()
    page.goto('http://localhost:5173')
    page.wait_for_load_state('networkidle')  # CRITICAL
    # ... test logic
    browser.close()

Reference Examples

If examples/ directory exists in project:

• examples/element_discovery.py — discovering buttons, links, inputs
• examples/static_html_automation.py — file:// URLs for local HTML
• examples/console_logging.py — capturing console logs

Rules

• ALWAYS wait for networkidle before inspecting DOM on dynamic apps
• NEVER read scripts/with_server.py source — use --help and invoke as black box
• Prefer MCP Playwright tools (mcp__plugin_playwright_playwright__*) over writing Python scripts when available
• Identify selectors from rendered state (recon-first), not from guessing
• Use descriptive selectors: text=, role=, CSS, or IDs — never positional selectors like nth-child
• Always close browser when done in Python scripts
• Re-check trace_path after tests pass to confirm full coverage