← Back to Skills Marketplace
422
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install playwright-browser
Description
Use Playwright to browse websites with a real (non-headless) browser and extract data by hooking network responses. Use when the user wants to: - View a webs...
Usage Guidance
This skill contains legitimate Playwright browser automation code, but also includes example scripts that will read from and write to your home/Desktop without that being declared. Before installing or running: 1) Review or remove scripts you don't trust (notably intercept_sina_images.py and parse_attendance.py). intercept_sina_images.py will save images to ~/Desktop/sina; parse_attendance.py will read ~/Desktop/考勤 and write an Excel file — run these only on a non-sensitive machine or in a sandbox. 2) Expect Playwright to download browser binaries when you run 'playwright install chromium'. 3) Ask the publisher why an attendance parser (parse_attendance.py) is bundled with a browser-scraping skill; this looks unrelated. 4) If you must run it, run a targeted script (browser_agent.py) rather than running all included examples, and inspect any script that performs filesystem operations. 5) If you need higher assurance, run the code in an isolated VM or container and restrict its filesystem access.
Capability Analysis
Type: OpenClaw Skill
Name: playwright-browser
Version: 1.0.0
The skill bundle provides browser automation via Playwright but contains several significant security vulnerabilities and risky configurations. Specifically, `scripts/browser_agent.py` is vulnerable to JavaScript injection because it uses unsanitized string interpolation within `page.evaluate()` calls (e.g., in `find_links_by_text`, `search_page_content`, and `focus_on_element`). Furthermore, the browser is launched with flags that explicitly disable critical Chromium security features, such as `--disable-features=IsolateOrigins,site-per-process`, which weakens the browser's sandbox. While the included utility scripts for scraping Sina and 12306 appear to serve their stated purpose, the underlying implementation flaws pose a risk if the agent navigates to untrusted or attacker-controlled websites.
Capability Assessment
Purpose & Capability
The SKILL.md describes non-headless Playwright browsing and network hooking, which matches browser_agent.py and the 12306/test scripts. However, the bundle also contains scripts that read user files and write to the user's Desktop (intercept_sina_images.py saves images to ~/Desktop/sina; parse_attendance.py reads ~/Desktop/考勤 and writes an Excel file). parse_attendance.py is unrelated to the stated browsing purpose. Those file-access behaviors are not declared in the skill metadata and are disproportionate to the described capability.
Instruction Scope
SKILL.md instructs installing Playwright and using the SyncBrowserAgent APIs; it does not warn that example scripts will read from or write to the user's home/Desktop. The code registers network response handlers and some example scripts automatically save binary responses to disk and read local directories — actions outside the SKILL.md's advertised scope and not restricted or highlighted in the runtime instructions.
Install Mechanism
No install spec in registry; SKILL.md instructs pip install playwright and playwright install chromium (standard for Playwright). There are no obscure URL downloads in the bundle. The Playwright browser installation will download browser binaries (expected).
Credentials
The skill declares no required environment variables or config paths, yet multiple scripts access the local filesystem under the user's home directory (e.g., SAVE_DIR = ~/Desktop/sina and INPUT_DIR = ~/Desktop/考勤). This file I/O is not reflected in metadata and can expose or modify user data without an explicit declared permission or prompt.
Persistence & Privilege
The skill is not always-enabled and does not request elevation or modify other skills' configs. It runs code only when invoked; there is no declared persistent/auto-install behavior in the manifest.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install playwright-browser - After installation, invoke the skill by name or use
/playwright-browser - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release of playwright-browser skill.
- Launches a real (non-headless) Chrome/Chromium browser to navigate and interact with websites.
- Can scrape rendered DOM content and extract data from JavaScript-heavy or SPA sites.
- Hooks network responses, allowing capture of XHR/Fetch API calls.
- Supports finding and clicking links, as well as searching for keywords within web pages.
- Provides examples and safety guidelines for secure use.
Metadata
Frequently Asked Questions
What is playwright-browser?
Use Playwright to browse websites with a real (non-headless) browser and extract data by hooking network responses. Use when the user wants to: - View a webs... It is an AI Agent Skill for Claude Code / OpenClaw, with 422 downloads so far.
How do I install playwright-browser?
Run "/install playwright-browser" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is playwright-browser free?
Yes, playwright-browser is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does playwright-browser support?
playwright-browser is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created playwright-browser?
It is built and maintained by elijahxb (@elijahxb); the current version is v1.0.0.
More Skills