← Back to Skills Marketplace
wang
by
mingxiangmiu
· GitHub ↗
· v1.0.0
· MIT-0
109
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install miu
Description
A fast Rust-based headless browser automation CLI with Node.js fallback that enables AI agents to navigate, click, type, and snapshot pages via structured co...
Usage Guidance
This skill is a documentation wrapper for the agent-browser CLI and is coherent with that purpose. Before installing or running agent-browser (npm install -g agent-browser or git clone), verify the upstream package/repo (the docs reference https://github.com/vercel-labs/agent-browser) to ensure you trust it. Be cautious with commands that save/load session state (auth.json), upload local files, intercept network requests, or eval JavaScript — those operations can expose cookies, auth tokens, or local files. Run the CLI in an isolated environment (container or sandbox) if you need to evaluate it, and avoid loading saved states or running commands that touch sensitive files from untrusted sources.
Capability Analysis
Type: OpenClaw Skill
Name: miu
Version: 1.0.0
The skill bundle provides a comprehensive wrapper for the 'agent-browser' CLI, granting the agent high-risk capabilities such as accessing browser cookies and localStorage, intercepting network traffic, and executing arbitrary JavaScript via the 'eval' command. While these features are aligned with the stated purpose of browser automation, they constitute meaningful high-risk behaviors that could be exploited for session hijacking or data exfiltration. Additionally, the metadata in '_meta.json' contains future-dated timestamps (2026) and inconsistent owner IDs, which is unusual, though no explicit malicious payloads or exfiltration endpoints were identified in SKILL.md.
Capability Assessment
Purpose & Capability
Name/description (browser automation) align with the SKILL.md commands and required binaries. Requesting node and npm makes sense because the skill documents installing a Node CLI (agent-browser) and references a GitHub repo. Nothing requested in metadata appears unrelated to browser automation.
Instruction Scope
SKILL.md stays on-topic (commands for navigation, snapshot, interactions, cookies, storage, network routing, JS eval, state save/load, file upload). These capabilities are expected for a headless browser wrapper, but they legitimately allow reading/writing session state (auth.json), uploading local files, intercepting/mocking network requests, and executing JS in page context — all of which can expose sensitive data if used improperly. The instructions do not themselves instruct exfiltration, but they enable powerful actions that an agent could perform.
Install Mechanism
The skill is instruction-only (no install spec), so it won't install code itself. The documentation tells users to npm install -g agent-browser or build from GitHub; that means any actual installation pulls code from npm or GitHub — normal for CLI tooling but you should vet the upstream package before running those install commands.
Credentials
The skill declares no environment variables or secrets and only requires node/npm binaries. That is proportionate for a Node-based CLI wrapper. No unrelated credentials or config paths are requested.
Persistence & Privilege
always:false and default autonomous invocation are set (normal). The skill does not request permanent system-wide presence or modify other skills' configs. Saving/loading state and file access are part of its functional scope but are not persistence privileges requested by the skill itself.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install miu - After installation, invoke the skill by name or use
/miu - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
- Initial release of Agent Browser skill (version 1.0.0).
- Enables fast, Rust-based headless browser automation via CLI with Node.js fallback.
- Supports navigation, interaction (click, type, form fill), structured page snapshots, and information extraction.
- Includes commands for screenshots, PDF generation, video recording, network/request control, browser/session state management, and more.
- Designed for web automation, data extraction, programmatic form handling, and UI testing.
Metadata
Frequently Asked Questions
What is wang?
A fast Rust-based headless browser automation CLI with Node.js fallback that enables AI agents to navigate, click, type, and snapshot pages via structured co... It is an AI Agent Skill for Claude Code / OpenClaw, with 109 downloads so far.
How do I install wang?
Run "/install miu" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is wang free?
Yes, wang is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does wang support?
wang is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created wang?
It is built and maintained by mingxiangmiu (@mingxiangmiu); the current version is v1.0.0.
More Skills