المهارات browser-act
🌐

browser-act

مخاطر منخفضة ⚙️ الأوامر الخارجية🌐 الوصول إلى الشبكة📁 الوصول إلى نظام الملفات

Automate web browsers with stealth mode and captcha solving

AI agents need reliable browser automation for web scraping, form filling, and data extraction. browser-act provides a structured CLI with anti-detection capabilities and built-in captcha solving.

يدعم: Claude Codex Code(CC)
🥉 74 برونزي
1

تنزيل ZIP المهارة

2

رفع في Claude

اذهب إلى Settings → Capabilities → Skills → Upload skill

3

فعّل وابدأ الاستخدام

اختبرها

استخدام "browser-act". browser-act stealth-extract https://news.ycombinator.com

النتيجة المتوقعة:

Returns page content as markdown with titles, scores, and comments extracted

استخدام "browser-act". browser-act state

النتيجة المتوقعة:

Returns indexed list: [1] input 'Search', [3] button 'Submit', [5] link 'About'

استخدام "browser-act". browser-act navigate https://example.com && browser-act wait stable && browser-act screenshot

النتيجة المتوقعة:

Opens page, waits for stability, returns screenshot path

التدقيق الأمني

مخاطر منخفضة
v1 • 4/28/2026

Static analysis detected 283 potential issues, primarily false positives from markdown documentation of CLI commands. The skill uses a structured CLI interface for browser automation with transparent security documentation. All external command patterns are well-defined CLI invocations, not arbitrary shell execution. Network access is limited to documented cloud services for captcha solving. Filesystem access is scoped to browser profile directories. Human oversight mechanisms (policies, human-assist-url) provide appropriate safeguards.

6
الملفات التي تم فحصها
694
الأسطر التي تم تحليلها
3
النتائج
1
إجمالي عمليات التدقيق

عوامل الخطر

⚙️ الأوامر الخارجية (1)
🌐 الوصول إلى الشبكة (1)
📁 الوصول إلى نظام الملفات (1)
تم تدقيقه بواسطة: claude

درجة الجودة

50
الهندسة المعمارية
100
قابلية الصيانة
87
المحتوى
50
المجتمع
90
الأمان
83
الامتثال للمواصفات

ماذا يمكنك بناءه

Extract product data from e-commerce sites

Automate scraping product listings, prices, and reviews from websites with anti-scraping protection. Use stealth mode to avoid detection.

Fill forms and submit data automatically

Navigate multi-step forms, handle dropdowns and checkboxes, and submit data. Credential-based logins require human confirmation first.

Monitor changing web content

Take periodic screenshots or extract text to detect changes on monitored pages. Use parallel sessions to check multiple sites.

جرّب هذه الموجهات

Simple page extraction
Extract the main content from {url} and return it as markdown
Form filling workflow
Open {url} in a new browser, find the login form, wait for it to load, and show me what fields are available
Multi-step automation
Navigate to {url}, click the element labeled '{element}', wait for the page to stabilize, then extract all headings
Captcha handling
Open {url}, attempt to solve any captcha present, then continue with the login flow

أفضل الممارسات

  • Run wait stable and state after any action that changes the page content
  • Use stealth-extract for simple content retrieval without session overhead
  • Check references/policies.md at the start of every task for human assist triggers

تجنب

  • Do not fill credentials unless explicitly provided by the user
  • Do not chain commands that require parsing intermediate output
  • Do not assume element indices remain valid after page navigation

الأسئلة المتكررة

How do I install browser-act?
Run: uv tool install browser-act-cli --python 3.12
What is the difference between Stealth and Real Chrome mode?
Stealth uses anti-detection fingerprinting for bot-protected sites. Real Chrome connects to your local Chrome with existing login sessions.
How do I handle captchas?
Use browser-act solve-captcha when a captcha challenge appears on the page.
Can I run multiple browser sessions in parallel?
Yes, use --session flag to create isolated sessions: browser-act --session task-a browser open <id> <url>
Where are browser profiles stored?
macOS: ~/Library/Application Support/browseract/, Windows: %APPDATA%\browseract, Linux: ~/.local/share/browseract
How do I stop automation and get human help?
Policies trigger human assist automatically for credentials, failed captchas, payments, and stuck operations. Use human-assist-url to request remote access.

تفاصيل المطور

المؤلف

browser-act

الترخيص

MIT

مرجع

main

بنية الملفات