Usage Scenarios

Scenario Overview

Scenario	Feasibility	Notes
1. Google Maps Bulk Scraping	✅ Yes	Sidebar scroll feature enables full data collection
2. KakaoTalk Simple Auth	✅ Yes	Can click “Request Auth” button (mobile approval required by user)
3. DART Report Analysis	✅ Yes	Can extract text from web viewer
4. Certificate Login (Korea)	❌ No	Cannot control external processes (EXE/ActiveX)

---

Smart Prompts

Copy-paste prompts that handle complex tasks without requiring CSS selector knowledge. Just paste these directly to Claude.

Google Maps Business Scraping — Universal Prompt

Copy and use directly

Google Maps class names change frequently. This prompt instructs Claude to analyze the current page structure and find the correct selectors automatically.

I have Google Maps search results open.
First, use execute_script to analyze the page structure and identify:
- The business list container (usually a div with role="feed")
- Selectors for business name, rating, review count, and address

Then:
1. Use scroll_page with the selector option to scroll only the sidebar (not the whole page)
2. After each scroll, use wait_for_element to wait for new items to load
3. Repeat scroll → wait → extract until at least 30 businesses are collected
4. Save the collected data to google_maps_results.csv

Important: Google Maps uses dynamically generated class names (e.g., .qBF1Pd, .MW4etd).
Always verify the current selectors with execute_script before scraping.

Naver Maps Business Scraping — Universal Prompt

I have Naver Maps search results open.
Use execute_script to analyze the page structure and identify the business list selectors.

Naver Maps typically uses:
- List container: #_pcmap_list_scroll_container or similar ID
- Business card: .CHC5F or similar class
- Business name: .TYaxT or similar class

After analysis:
1. Use scroll_page(selector: "container_selector", direction: "down") on the scroll container
2. Use wait_for_element to wait for new items to load
3. Repeat until 30+ items are collected
4. Save to naver_maps_results.csv

Selectors may differ from expected — verify page structure first.

Generic List Page — Universal Prompt

I want to scrape repeating list items from the current page.

First, use execute_script to analyze the page and find:
1. The common parent container selector for repeating items
2. Selectors for title, description, link within each item

Then use extract_structured_data with:
- containerSelector: selector for individual items
- fields: { title: "title_selector", description: "desc_selector", link: "link_selector" }

If needed, scroll to load more items and save to CSV.

---

1. Google Maps Bulk Scraping

The most powerful use case. If logged in, you can also scrape personalized search results.

Goal

Scrape 50 “Gangnam cafe” search results with scrolling and save to cafes.csv.

Example Prompt

Search for "Gangnam cafe" on Google Maps (google.com/maps).
Then scroll the left sidebar (div[role="feed"]) and extract:
- Business name (.qBF1Pd)
- Rating (.MW4etd)
- Address (.W4Efsd)
Collect 50 items and save to gangnam_cafes.csv.

Technical Notes

• Container-specific scrolling: Use the selector option in scroll_page to scroll only the sidebar.
• Selector: Google Maps list container typically has div[role="feed"] attribute.

Google Maps Selector Warning

CSS class names like .qBF1Pd, .MW4etd change between builds. Use the Universal Prompt above to let Claude detect current class names.

---

2. KakaoTalk Simple Authentication

Useful for government sites (Korea) that require simple authentication.

Goal

Enter name, birth date, phone number on the login page and click “Request Authentication”.

Example Prompt

On the current tax office login page, select KakaoTalk simple authentication.
Enter name: "Hong Gildong", birth date: "19900101", phone: "01012345678"
Check "Agree to all" and click the "Request Authentication" button.

Notes

• Mobile approval required: PC browser can only request authentication. Actual approval must be done on the smartphone KakaoTalk app.
• CAPTCHA: If security images appear, automation becomes difficult.

---

3. DART Report Analysis (Korea)

Let AI read massive reports from the Financial Supervisory Service electronic disclosure system (dart.fss.or.kr).

Goal

Open Samsung Electronics business report and summarize the “Management Discussion & Analysis” chapter.

Example Prompt

Open Samsung Electronics' latest business report on DART.
Click "IV. Management Discussion & Analysis" in the left table of contents.
Read the content using read_browser_content and summarize the key points.

Technical Notes

• iFrame handling: DART viewer may have content inside an iframe. If read_browser_content can’t access iframe content, navigate directly to the frame’s URL using navigate_to.

---

4. Certificate Login (Limitations)

Why It’s Not Possible

Korean legacy certificate (공동인증서) windows are not HTML rendered by Chrome. They are Windows application windows (EXE) launched by security plugins like AnySign4PC.

BrowseHand can only control browser DOM — it cannot click or type into windows outside the browser.

Alternatives

• Browser certificates: If the site supports cloud-based browser certificates instead of local disk certificates, it may work.
• Simple authentication: KakaoTalk, PASS, Naver authentication use web-based UI and are controllable. (See Scenario 2)