← 返回 Skills 市场
Wayne Agent Browser
作者
WechatGpt798
· GitHub ↗
· v1.0.0
· MIT-0
97
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install wayne-agent-browser
功能描述
Headless browser automation CLI optimized for AI agents with accessibility tree snapshots and ref-based element selection
安全使用建议
This skill appears to be a coherent browser-automation CLI, but exercise caution before installing or running it. Things to do before proceeding: 1) Verify the npm package and GitHub repo (https://github.com/vercel-labs/agent-browser) are legitimate and match the version you expect. 2) Prefer installing and running the CLI in an isolated environment (container or VM) because `npm install -g` runs package install scripts and `agent-browser install` downloads Chromium binaries. 3) Avoid loading or saving auth state files (cookies/localStorage) unless you trust the environment and repository; such files can contain sensitive tokens. 4) If you must use it, restrict its filesystem/network permissions and review any downloaded binaries' checksums. 5) If you want lower risk, ask the skill author to declare required binaries and provide an explicit, vetted install spec (trusted release URLs/checksums) so you can audit the installation sources.
功能分析
Type: OpenClaw Skill
Name: wayne-agent-browser
Version: 1.0.0
The skill bundle provides documentation and usage instructions for the 'agent-browser' CLI, a legitimate headless browser automation tool developed by Vercel Labs. The instructions in SKILL.md are purely functional, covering standard automation tasks such as navigation, element interaction, and session management without any evidence of malicious intent, data exfiltration, or prompt injection.
能力评估
Purpose & Capability
The SKILL.md clearly targets a CLI named `agent-browser` and describes installing it with `npm install -g agent-browser`, downloading Chromium, and running many CLI commands. However the registry metadata lists no required binaries or install spec. That mismatch (the skill essentially requires a third-party CLI and npm installs, but the package metadata declares none) is inconsistent and worth noting.
Instruction Scope
The instructions focus on browser automation (navigation, snapshots, refs) which is consistent with the description. However they explicitly instruct saving/loading auth state files (e.g., `state save auth.json` / `state load auth.json`), controlling network routing/mocking, and installing runtime dependencies. Those file I/O and network-control steps can be used to persist or exfiltrate sensitive tokens/cookies or to intercept/modify requests — while legitimate for browser automation, they expand the scope beyond simple read-only scraping and should be considered when granting the tool access.
Install Mechanism
There is no install spec in the registry, but SKILL.md tells users to run `npm install -g agent-browser` and `agent-browser install` (which downloads Chromium). Global npm installs execute package install scripts and will run code from the npm package; the Chromium download will pull external binaries. This is a higher-risk install pattern unless you verify the npm package and download sources are trustworthy (the homepage points to a GitHub repo which helps, but the skill metadata and registry did not declare these installs).
Credentials
The skill declares no required environment variables or credentials (registry shows none). SKILL.md mentions an optional AGENT_BROWSER_SESSION env var for convenience, but no secrets are requested by the skill itself. The main proportionality concern is that the tool reads/writes auth state files (cookies/localStorage), which may contain sensitive credentials even though no env secrets are required.
Persistence & Privilege
The skill is not always-enabled and does not request elevated platform privileges. However the recommended workflows include persisting browser state to disk and loading it later (state save/load), so the tool can create long-lived local files containing session cookies/storage. This is normal for a browser automation tool but increases persistence of sensitive tokens on disk.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install wayne-agent-browser - 安装完成后，直接呼叫该 Skill 的名称或使用
/wayne-agent-browser触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.0.0
Initial release of agent-browser skill.
- Headless browser automation CLI specialized for AI agents.
- Deterministic, ref-based element selection using accessibility tree snapshots.
- Supports multi-step workflows, session isolation, and state persistence.
- Comprehensive set of commands for navigation, interaction, state check, network & storage control.
- Includes best practices and usage examples for efficient automation.
元数据
常见问题
Wayne Agent Browser 是什么？
Headless browser automation CLI optimized for AI agents with accessibility tree snapshots and ref-based element selection. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 97 次。
如何安装 Wayne Agent Browser？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install wayne-agent-browser」即可一键安装，无需额外配置。
Wayne Agent Browser 是免费的吗？
是的，Wayne Agent Browser 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Wayne Agent Browser 支持哪些平台？
Wayne Agent Browser 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Wayne Agent Browser？
由 WechatGpt798（@wechatgpt798）开发并维护，当前版本 v1.0.0。
推荐 Skills