基于微软 Playwright CLI 的浏览器自动化技能。用于 Web 测试、页面交互、自动化工作流。当用户提到 playwright、browser automation、Web 测试、E2E 测试时触发。
Browser
Obclip
Try itInstall, verify, and operate the obclip CLI to clip live web pages into Markdown or Obsidian notes. Use when Codex needs to install the @harris7/obclip npm p...
What it does
Install, verify, and operate the obclip CLI to clip live web pages into Markdown or Obsidian notes. Use when Codex needs to install the @harris7/obclip npm p...
The skill document
Obclip
Use this skill to make obclip work predictably instead of trial-and-error.
Workflow
- Verify how
obclipshould be invoked. - Pick the browser mode that matches the target site.
- Run the smallest command that can succeed.
- Escalate waiting and browser state only when the output is incomplete.
- Report the final saved path or the concrete failure reason.
1. Verify Invocation
- Prefer
obclipif it is already onPATH. - If
obclipis missing, install the npm package from install-and-invoke.md. - If you are working inside the source repo and the global command is unavailable, fall back to
node \\dist\\cli.cjs .... - After installation, verify with
obclip --helpornpx @harris7/obclip --help.
2. Pick Browser Mode
- Use the default bundled Playwright Chromium for public pages that do not need login state.
- Use
--browser-executable ""when the user already has a dedicated Chromium build and wants a fixed browser binary. - Use
--browser-profile ""for sites that require login state, cookies, or a persistent session. - Use
--headfulwhen the user needs to see the browser or perform the initial login into a fresh profile. - Do not point
--browser-profileat the user's normal daily Chrome profile. Prefer a dedicated profile directory.
3. Start With The Smallest Command
- Default public-page capture:
obclip --output "\\" - Logged-in or anti-bot page: add
--browser-executable,--browser-profile, and usually--headful - If the user wants Markdown on stdout, omit
--output - If the user wants the note opened in Obsidian, add
--openand the optional vault/URI flags
Use the concrete command templates in command-recipes.md.
4. Escalate Waiting In Order
- First run without custom waits when the page is simple.
- If the page is incomplete but visibly loading dynamic content, add
--settle-ms 3000or--settle-ms 5000. - If the page is an SPA and the content mounts later, add
--wait-selector "". - Prefer broad, content-bearing selectors such as
article,main,[role="main"], or a stable post container. Avoid selectors that only match nav, login shells, or skeleton loaders. - If the page still shows a login wall, fix browser state with
--browser-profile; extra delay does not solve missing authentication.
Use troubleshooting.md when the first pass fails.
5. Interpret Results Correctly
- Successful file saves print
Saved note:tostderr. - Failures print a direct reason such as save failure, invalid browser path, or wait timeout.
- Keep
stdoutclean unless the user explicitly wants the Markdown body in the terminal or piped into another command.
6. Windows Path Rules
- Quote Windows paths that may contain spaces.
- For
--output, prefer an existing directory or a path ending in\\or/when you want directory semantics. - Example:
--output "D:\\data\\Clippings\\"
References
- Installation and invocation: install-and-invoke.md
- Command templates: command-recipes.md
- Failure handling: troubleshooting.md
Related skills
Use clip2md to configure an access token, save web pages or article URLs as Markdown clipping tasks, check remaining daily/permanent quota, query task status, and wait for clipping completion or failure. Use when the user mentions clip2md, 剪藏, clip, 保存网页为 Markdown, 查询额度, 提交链接, 查询任务, 等待剪藏完成, or wants
OpenClaw web clipper skill: save any web page as clean local Markdown with YAML frontmatter. Dual-engine extraction (readability-lxml + trafilatura fallback chain), Chinese filename safe, batch URL clipping with dedup, output lands in knowledge/clippings/ ready for kb-retriever indexing. Use when user asks to clip/save/collect a web page or article (剪藏/收藏/保存网页 文章/网页转 Markdown). 中文:OpenClaw 网页剪藏工具。把任意网页保存为带 frontmatter 的本地 Markdown:双引擎正文提取(readability-lxml + trafilatura 降级链)、中文文件名安全、批量剪藏 + 去重;输出直通 knowledge/clippings/, 配合 kb-retriever 建索引即可检索,构成六件套的「输入」环节(家 initializer → 内容 memory-distill → 状态 tracker → 知识 kb-retriever → 健康 auditor → 输入 web-clipper)。
Thin OpenClaw and ClawHub wrapper for the published clawfetch npm CLI, used to fetch web pages, GitHub READMEs, and Reddit threads as markdown.
Stealth browser automation using CloakBrowser to bypass bot detection on protected sites. Use when web_fetch fails with anti-bot blocks, CAPTCHA, or access denied errors. Use for scraping reviews, protected pages, or any site that blocks standard Playwright/Puppeteer automation. Not for sites requiring login credentials or solving interactive CAPTCHAs.
Web extraction for LLMs and agents. Scrape, crawl, map, search, extract, summarize, diff, monitor, and research any URL into clean Markdown, text, or JSON, i...