cuj-screenshots（cuj 截图）

CUJ Screenshots Skill

Automate visual documentation of web app user journeys using headless browser automation.

When to Use This

Always run CUJ captures after UI changes. This is part of the dev loop, not just documentation.

Triggered By:

• Any frontend component change
• CSS/styling updates
• New features that affect user flow
• Bug fixes that change visible behavior
• Before sharing work with humans ("here's what it looks like now")

The Loop:

Make UI change → Run CUJs → Review screenshots →
  ├─ Looks good? → Update GIFs, commit, share link
  └─ Looks wrong? → Fix bug, repeat

Proactive Use:

When finishing UI work, don't just say "done" — capture CUJs, update the GIFs, and share:

"Updated the task card styling. Here's the new CUJ: https://github.com/user/repo/blob/main/docs/CUJs.md"

This gives your human instant visual feedback without them needing to run the app.

What It Does

1. Launches headless Chromium via Playwright
2. Navigates through a defined user journey
3. Captures screenshots at each step
4. Stitches screenshots into an animated GIF
5. Outputs to a docs folder for commit

Requirements

• Playwright: uv run --with playwright python script.py
• Chromium: Auto-installed by Playwright on first run
• ImageMagick: For GIF creation (convert command)

Install Playwright browsers (one-time):

uv run --with playwright playwright install chromium

Quick Start

1. Define Your CUJ

Create a Python script for each journey:

import asyncio
from playwright.async_api import async_playwright
import os

async def capture_my_cuj():
    output_dir = "/tmp/cuj-screenshots/my-cuj"
    os.makedirs(output_dir, exist_ok=True)

    async with async_playwright() as p:
        browser = await p.chromium.launch(headless=True)
        context = await browser.new_context(
            viewport={"width": 390, "height": 844},  # Mobile
            device_scale_factor=2  # Retina
        )
        page = await context.new_page()

        # Step 1
        await page.goto("http://localhost:5173")
        await page.wait_for_timeout(1500)
        await page.screenshot(path=f"{output_dir}/01-initial.png")

        # Step 2: Interact
        await page.click("button:has-text('Submit')")
        await page.wait_for_timeout(1000)
        await page.screenshot(path=f"{output_dir}/02-submitted.png")

        # ... more steps

        await browser.close()

if __name__ == "__main__":
    asyncio.run(capture_my_cuj())

2. Run Capture

uv run --with playwright python my_cuj.py

3. Create GIF

convert -delay 150 -loop 0 /tmp/cuj-screenshots/my-cuj/*.png output.gif

• -delay 150 = 1.5 seconds per frame (hundredths of a second)
• -loop 0 = infinite loop

4. Commit & Share

mv output.gif docs/gifs/my-cuj.gif
git add docs/gifs/my-cuj.gif docs/CUJs.md
git commit -m "docs: add my-cuj GIF"
git push

Viewport Presets

Common Interactions

# Click button by text
await page.click("button:has-text('Submit')")

# Fill input
await page.fill("input[name='email']", "test@example.com")

# Select dropdown
await page.select_option("select#plan", "premium")

# Wait for element
await page.wait_for_selector(".success-message")

# Scroll
await page.evaluate("window.scrollBy(0, 300)")

# Keyboard shortcut
await page.keyboard.press("Escape")

Full Workflow Example

See scripts/capture-cujs.py for a complete example that:

• Starts backend + frontend servers
• Captures multiple CUJs
• Generates GIFs
• Updates CUJs.md

Tips

• Wait times: Use wait_for_timeout(1000) after interactions for animations
• Selectors: Prefer text= or has-text() over fragile CSS selectors
• Error handling: Wrap interactions in try/except, screenshot on error
• Naming: Use numbered prefixes (01-, 02-) for sort order
• GIF size: Keep under 500KB for GitHub README display

CUJs.md Documentation

Create a docs/CUJs.md file to document each journey with embedded GIFs:

# Critical User Journeys (CUJs)

## CUJ 1: App Tour

**Goal:** Navigate the main app layout.

**Steps:**
1. Open app → Activity tab
2. Switch to Live tab
3. Return to Activity

![App Tour](./gifs/cuj1-app-tour.gif)

---

## CUJ 2: Create Task

**Goal:** Create a new plan and task.

**Steps:**
1. Click "+ Plan"
2. Enter name, submit
3. Click "+ Task"
4. Enter name, submit

![Create Task](./gifs/cuj2-create-task.gif)

This serves as both documentation and visual regression baseline.

Verifying UI Changes

After making UI changes:

1. Run the CUJ capture scripts
2. Review the new screenshots (you can read image files!)
3. If they look correct, update GIFs and commit
4. Update CUJs.md if steps changed
5. If something's wrong, you caught a bug!

Full Workflow

# 1. Make UI changes
# 2. Start servers
# 3. Capture CUJs
uv run --with playwright python scripts/capture-cujs.py

# 4. Review screenshots (agent can view these)
# 5. Create GIFs
convert -delay 150 -loop 0 /tmp/cuj-screenshots/cuj1/*.png docs/gifs/cuj1.gif

# 6. Update CUJs.md with new steps/descriptions
# 7. Commit everything
git add docs/CUJs.md docs/gifs/
git commit -m "docs: update CUJ screenshots after UI changes"
git push