Browser Orchestrator

Trigger Words

小蔥技能站7w4.net，專業的AI技能分享平臺。

browser automation, web scraping, data collection, web testing, UI testing, RPA, auto form fill, screenshot comparison, browser control, crawler, web interaction, anti-detection, headless browser

Capabilities

Five browser automation capabilities in one, from simple clicks to complex crawlers.

Capability 1: Web Data Scraping

Scraping Pipeline

1. Target Analysis — URL structure, page type, data location, anti-scrape measures
2. Strategy Selection — API interception vs DOM parsing vs visual recognition
3. Selective Extraction — Only fetch needed fields, avoid over-requesting
4. Data Cleaning — Dedup/format/type conversion
5. Structured Output — CSV/JSON/Database/direct file write

Anti-Anti-Scrape Strategies

• Randomized request intervals (human reading speed simulation)
• User-Agent rotation
• Cookie/Session persistence
• CAPTCHA solving (OCR integration/third-party services)
• IP rotation (proxy pool)
• Behavior humanization (random scrolls/mouse movement/dwell times)

Capability 2: Web Interaction Automation

• Form Filling — Auto-detect field types and fill (text/select/date/file upload)
• Click Navigation — Links/buttons/Tabs/menus/dropdowns
• Keyboard Operations — Shortcuts/text input/combo keys
• Drag & Drop — Sort/slider verification/file drag
• Popup Handling — Alert/Confirm/modal/new window
• File Operations — Upload/download/save

Capability 3: UI Testing & Regression

• Visual Regression — Page screenshots + pixel-level diff comparison
• Element Assertions — Element exists/visible/clickable/text content validation
• Flow Recording — Record steps into replayable scripts
• Multi-Browser Testing — Chrome/Firefox/Safari/Edge compatibility
• Responsive Testing — Layout verification at multiple resolutions
• Performance Metrics — Load time/FCP/Core Web Vitals

Capability 4: Anti-Detection Strategies

Capability 5: Multi-Instance Management

• Multi-tab Parallelism — Operate multiple pages simultaneously
• Multi-Browser Sessions — Isolated context environments
• Task Queues — Serial/parallel execution modes
• State Persistence — Session save/restore
• Resource Monitoring — Memory/CPU/network usage

Workflows

Mode A: Data Collection Task

"Scrape XX data from XX website" → Analyze page → Plan strategy → Scrape → Clean → Deliver

Mode B: Automation Workflow

"Automate XX operation daily" → Record/write script → Schedule execute → Result notification

Mode C: UI Testing

"Check if XX page looks correct at different resolutions" → Multi-res screenshots → Diff comparison → Issue report

Mode D: Anti-Detection Crawling

"XX site has anti-scrape measures, bypass them" → Analyze defenses → Build counter-strategy → Stable collection

Output Spec

## Browser Automation Task Report
- **Task Type**: Scrape/Interact/Test/Other
- **Target URL**: ...
- **Steps Executed**: N steps
- **Data Volume**: N records / N files
- **Elapsed Time**: X:XX
- **Exceptions/Retries**: N times
- **Output Files**: <path list>

Security & Compliance

• Respect robots.txt and website ToS
• Reasonable request rates, don't overwhelm target servers
• Don't scrape personal data without explicit authorization
• Use data only for stated purposes

個人創作者計費通道上線後直接升級計費

文件寫得詳細全面，涵蓋了瀏覽器自動化的各個方面，結構清晰看起來很專業。但這只是使用說明，沒有實際可執行的程式碼。相當於買了一本菜譜，卻沒有食材。真正想用的話，還得自己從頭寫程式碼。內容不錯但實用性不足。