القائمة
When you need to interact with web pages using the user's logged-in browser → control Chrome via @playwright/cli daemon, reusing cookies/session with persistent tabs.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
When prompt, tool, or agent logic changes need regression testing → run eval benchmarks to detect regressions or improvements.
When a task can be parallelized (bulk code review, multi-module refactor, batch analysis) → split into subtasks running in isolated git worktrees via multiple workers.
When the user wants to research academic papers → search ArXiv, extract PDFs, summarize/review/Q&A, generate experiment code.
When user needs AnyGenIO content generation → unified CLI for task execution with progressive disclosure.
When you need to persist files (reports, docs, evidence) beyond the session → create/query/delete durable artifacts.
When the user wants text-to-speech audio → generate voice files locally (macOS say + afconvert, outputs m4a).
| name | browser-use |
| description | When you need to interact with web pages using the user's logged-in browser → control Chrome via @playwright/cli daemon, reusing cookies/session with persistent tabs. |
| triggers | {"intent_patterns":["浏览器|browser|网页|webpage|打开网站|open.*url|截图|screenshot","x\\.com|twitter|小红书|xiaohongshu|redbook|微博|weibo|抖音|douyin|tiktok","reddit|hacker.?news|github\\.com|youtube|bilibili|B站|知乎|zhihu|豆瓣|douban","instagram|linkedin|facebook|threads|bluesky|mastodon|telegram","登录.*页面|login.*page|填写表单|fill.*form|submit.*form|提交表单","看看这个网站|帮我打开|check.*this.*site|visit.*page|go to","页面.*内容|page.*content|抓取|scrape|爬取|extract.*from.*page","点击|click|输入|type.*into|scroll|滚动|翻页|next.*page","网站.*截图|capture.*page|保存.*网页|save.*page","帮我.*操作|automate.*web|自动化.*网页|web.*automation","查看.*网页|look.*at|看一下|帮我看看|打开看看","发.*推|post.*tweet|发.*帖|create.*post|点赞|like|转发|retweet|repost","刷.*首页|browse.*feed|看.*热搜|trending|搜索.*用户|search.*user","评论|comment|关注|follow|收藏|bookmark|分享|share","下单|checkout|加购|add.*cart|购物|shopping","复用.*tab|reuse.*tab|之前.*标签页|切换.*标签"],"context_signals":{"keywords":["浏览器","browser","网页","截图","打开","navigate","snapshot","tab","标签页","页面","网站","click","点击","form","表单","scrape","抓取","小红书","微博","知乎","x.com","twitter","发帖","评论","点赞"]},"confidence_threshold":0.6} |
| priority | 6 |
| requires_tools | ["bash"] |
| max_tokens | 4000 |
| cooldown | 10 |
通过 @playwright/cli daemon 连接用户真实 Chrome 浏览器（extension 模式），复用已登录的 session。
Tab 跨调用持久化 — daemon 进程常驻，tab、cookies、登录状态在多次调用间存活。
.env 中配置了 ALEX_BROWSER_BRIDGE_TOKEN（从扩展弹窗复制）open --extension），后续所有调用复用同一会话snapshot 获取页面无障碍树，每个元素有 ref（如 e44），用 click/fill/type 操作该 ref.playwright-cli/ 目录），不灌入 stdout# 首次打开 — 连接 Chrome 扩展（自动启动 daemon）
python3 skills/browser-use/run.py open
# 导航到 URL（自动确保 session 存在）
python3 skills/browser-use/run.py navigate --url https://x.com
# 获取页面快照（无障碍树 + 元素 ref）
python3 skills/browser-use/run.py snapshot
# 点击元素（ref 来自 snapshot）
python3 skills/browser-use/run.py click --ref e44
# 填充输入框（定位到 ref 再填入文本）
python3 skills/browser-use/run.py fill --ref e100 --text "hello world"
# 输入文本到当前焦点元素
python3 skills/browser-use/run.py type --text "search query" --submit true
# 悬停
python3 skills/browser-use/run.py hover --ref e50
# 下拉选择
python3 skills/browser-use/run.py select --ref e80 --value "option1"
# 截图
python3 skills/browser-use/run.py screenshot --filename page.png
python3 skills/browser-use/run.py screenshot --ref e44 --filename element.png
python3 skills/browser-use/run.py screenshot --full_page true --filename full.png
# 按键
python3 skills/browser-use/run.py press_key --key Enter
python3 skills/browser-use/run.py press_key --key "Control+a"
# 导航
python3 skills/browser-use/run.py go_back
python3 skills/browser-use/run.py go_forward
python3 skills/browser-use/run.py reload
# 执行 JavaScript
python3 skills/browser-use/run.py evaluate --function '() => document.title'
python3 skills/browser-use/run.py evaluate --function '(el) => el.textContent' --ref e44
# 执行 Playwright 代码
python3 skills/browser-use/run.py run_code --code 'async (page) => await page.title()'
# Tab 管理（跨调用持久！）
python3 skills/browser-use/run.py tabs # 列出所有 tab
python3 skills/browser-use/run.py tab_select --index 0 # 切换到 tab 0
python3 skills/browser-use/run.py tab_new --url https://... # 新开 tab
python3 skills/browser-use/run.py tab_close --index 1 # 关闭 tab
# 调试
python3 skills/browser-use/run.py console # 查看 console 日志
python3 skills/browser-use/run.py network # 查看网络请求
python3 skills/browser-use/run.py pdf # 导出 PDF
# 关闭会话
python3 skills/browser-use/run.py close
Tab 在 daemon 会话中持久化。跨调用复用示例：
# 调用 1: 打开 x.com
python3 skills/browser-use/run.py navigate --url https://x.com
# 调用 2: 直接 snapshot（还在 x.com 页面上！）
python3 skills/browser-use/run.py snapshot
# 调用 3: 操作 x.com 页面元素
python3 skills/browser-use/run.py click --ref e43
# 调用 4: 查看所有 tab
python3 skills/browser-use/run.py tabs
# 调用 5: 切换到另一个 tab
python3 skills/browser-use/run.py tab_select --index 0
# 打开 → 搜索 → 获取结果
python3 skills/browser-use/run.py navigate --url https://x.com/search
python3 skills/browser-use/run.py snapshot
# 找到搜索框 ref → 填入关键词
python3 skills/browser-use/run.py fill --ref SEARCH_REF --text "AI agent"
python3 skills/browser-use/run.py press_key --key Enter
python3 skills/browser-use/run.py snapshot
python3 skills/browser-use/run.py navigate --url https://x.com/compose/post
python3 skills/browser-use/run.py snapshot
python3 skills/browser-use/run.py fill --ref POST_REF --text "Hello world!"
python3 skills/browser-use/run.py click --ref SUBMIT_REF
python3 skills/browser-use/run.py navigate --url https://example.com/form
python3 skills/browser-use/run.py snapshot
# 逐一填充字段
python3 skills/browser-use/run.py fill --ref NAME_REF --text "张三"
python3 skills/browser-use/run.py fill --ref EMAIL_REF --text "test@example.com"
python3 skills/browser-use/run.py select --ref COUNTRY_REF --value "CN"
python3 skills/browser-use/run.py click --ref SUBMIT_REF
python3 skills/browser-use/run.py navigate --url https://www.xiaohongshu.com/explore
python3 skills/browser-use/run.py snapshot
# 滚动加载更多
python3 skills/browser-use/run.py evaluate --function '() => window.scrollBy(0, 800)'
python3 skills/browser-use/run.py snapshot
# 翻页
python3 skills/browser-use/run.py click --ref NEXT_PAGE_REF
snapshot 返回无障碍树，ref 是元素标识符（如 e44）