browser
Browser and device automation: pick the right engine (the host's preview panel first, an anti-detect browser over MCP for hard targets, CDP CLI or Playwright for test suites), test local web apps, cap
SKILL.mdskills/browser/SKILL.md
GitHubbrowser — automation engines, testing, and Android
Home branch in Memora:
root/skill/browser.Engine order: (1) the host's own preview panel —
HOSTS.md, verb Show app; (2) a real anti-detect browser driven over MCP; (3) Playwright or a CDP CLI for headless suites and CI. A local dev server is shown, not screenshotted, unless the task needs the pixels.
Choosing a part
| Task | Part |
|---|---|
| Choose the right browser engine for the task | parts/browser-automation/ |
| Standalone CDP CLI navigation, forms, extraction | parts/agent-browser/ |
| Anti-detect Camoufox browsing via BrowserMCP (logins, 2FA/TOTP, snapshots) | parts/browsermcp-automation/ |
| Production Playwright e2e suites, flaky-test repair, CI/CD | parts/playwright-pro/ |
| Interactive testing and console/network inspection of local web apps | parts/webapp-testing/ |
| Full-page / tall-page / SPA scroll-container screenshots via CDP | parts/full-page-screenshot/ |
| Validation-heavy scraping, crawling, PDF/table extraction pipelines | parts/universal-scraping-architect/ |
Drive an Android phone via USB (adb, UI dump, MIUI toggles, SMS inbox) |
parts/android-adb-automation/ |
Rules
- Walk
root/skill/browserin Memora before starting non-trivial automation; record new anti-bot or device findings as nodes. - Ask the owner when choosing between a logged-in session and a clean anti-detect profile, and before any action inside a signed-in account.
- Before any web deploy, show the running dev server in the host's panel.
- A project-specific capture recipe belongs inside that project, not in this router.
Related
research (skills/research/parts/searchmcp-research/ for multi-engine Camoufox research), reverse-engineering (skills/reverse-engineering/parts/browser-extension-reverse/), osint (platform collection).