OpenClaw Native Browser

A stable native browser (WKWebView) for OpenClaw agents. Opens a visible window with tab management, URL bar, and login helpers — every website works, includ...

This skill looks like a real native browser wrapper, but it asks you to download and pip-install code from an external GitHub repo and to modify your OpenClaw config and local files. Before installing: 1) review the repository source (https://github.com/yungookim/openclaw-browser) to confirm the code is trustworthy; 2) note that pip install -e . and setup scripts can execute arbitrary code — prefer running in an isolated environment or VM; 3) confirm you are on macOS (WKWebView) and that 'git' is available; 4) be cautious about providing account credentials to the login helpers (they persist cookies/sessions) — avoid reusing high-value credentials; 5) back up ~/.openclaw/openclaw.json before modifying it. If you cannot inspect the repo source or run it in isolation, consider this skill risky.

Type: OpenClaw Skill Name: openclaw-browser-2 Version: 0.1.0 The skill is classified as suspicious due to its broad capabilities that present significant security risks if misused or exploited via prompt injection. Key indicators include the `allowed-tools` declaration in `SKILL.md` granting arbitrary `Bash(python:*)` and `Bash(pip:*)` execution, the ability to execute arbitrary JavaScript (`skill.execute_js`) on any loaded webpage, and explicit functions for handling user credentials (`skill.login_perplexity`, etc.) and accessing/manipulating cookies (`skill.get_cookies`, `skill.set_cookie`). While these features align with the stated purpose of a native browser skill, they provide powerful primitives that could be leveraged for data exfiltration, unauthorized actions, or session hijacking if the agent is compromised, making it a high-risk component.