← 返回 Skills 市场

Ai Pa Browser Clawdbot

Name: Ai Pa Browser Clawdbot
Author: netanel-abergel

作者 Netanel Abergel · GitHub ↗ · v1.0.0 · MIT-0

cross-platform ⚠ suspicious

总下载

当前安装

版本数

在 OpenClaw 中安装

/install ai-pa-browser-clawdbot

功能描述

Headless browser automation CLI optimized for AI agents with accessibility tree snapshots and ref-based element selection

使用说明 (SKILL.md)

Agent Browser Skill

Fast browser automation via accessibility tree snapshots and ref-based element selection.

When to Use

Use agent-browser (this skill) when:

Automating multi-step web workflows
Need deterministic element selection
Working with complex SPAs or login flows
Need parallel sessions or video recording

Use built-in browser tool when:

Need screenshots/PDFs for visual analysis
Browser extension integration needed

Core Workflow

agent-browser open \x3Curl>
agent-browser snapshot -i --json   # Get interactive elements with refs
agent-browser click @e2
agent-browser fill @e3 "text"
agent-browser wait --load networkidle
agent-browser snapshot -i --json   # Re-snapshot after navigation

Most Common Commands

# Navigate
agent-browser open \x3Curl> | back | forward | reload | close

# Snapshot
agent-browser snapshot -i --json          # Always use these flags
agent-browser snapshot -s "#main" -i      # Scope to selector

# Interact (use @refs from snapshot)
agent-browser click @e1
agent-browser fill @e2 "text"
agent-browser press Enter
agent-browser select @e3 "value"
agent-browser scroll down 500

# Wait
agent-browser wait --load networkidle
agent-browser wait --text "Success"
agent-browser wait @e1

# State
agent-browser state save auth.json
agent-browser state load auth.json

Installation

npm install -g agent-browser
agent-browser install           # Download Chromium
agent-browser install --with-deps  # Linux: + system deps

Full Reference

For complete command list: read skills/ai-pa-browser-clawdbot/references/commands.md For examples: read skills/ai-pa-browser-clawdbot/references/examples.md

安全使用建议

This skill appears to be what it claims: a wrapper around an external 'agent-browser' CLI. Before installing or letting an agent use it, ensure you: (1) install agent-browser from a trusted source (verify the npm package and upstream repo), (2) avoid loading state files (auth.json) from untrusted origins, (3) be cautious about commands that set headers/credentials, upload files, or route/modify network requests (these can be used to exfiltrate data if misused), and (4) restrict the agent's permissions or review commands if the agent will run autonomously.

功能分析

Type: OpenClaw Skill Name: Developer: Version: Description: OpenClaw Agent Skill The skill bundle provides a wrapper for the 'agent-browser' CLI, a tool for headless browser automation. It includes several high-risk capabilities such as arbitrary JavaScript execution ('eval'), session state persistence ('state save/load'), and network request interception ('network route'). While these features are aligned with the stated purpose of enabling AI agents to automate web workflows, they represent a significant attack surface for session hijacking or code execution within the browser context. Additionally, the installation process ('agent-browser install') involves downloading external binaries (Chromium) and potentially modifying system dependencies, which are high-risk operations. No evidence of intentional malice or unauthorized data exfiltration was found in SKILL.md or the reference files.

能力评估

✓ Purpose & Capability

Name/description match the provided SKILL.md and reference docs: all commands (snapshots, refs, navigation, state save/load, network control, screenshots, recording) are expected features of a browser automation CLI.

✓ Instruction Scope

SKILL.md only instructs the agent to use the external 'agent-browser' CLI and documents its commands; it does not direct the agent to read unrelated system files or env vars, nor to send captured data to unexpected endpoints.

✓ Install Mechanism

This is an instruction-only skill with no install spec. The SKILL.md suggests installing via 'npm install -g agent-browser' (a normal distribution method); nothing in the bundle performs arbitrary downloads or writes to disk.

ℹ Credentials

The skill itself requests no environment variables or credentials (requires none). However, the documented CLI supports saving/loading state (auth.json), setting credentials/headers, uploading files, and network routing — all normal for a browser tool but potential vectors for leaking sensitive data if the agent is instructed to save or transmit secrets.

✓ Persistence & Privilege

always is false and the skill has no install hooks or behavior that modifies other skills or global agent settings. It does not request permanent platform presence.

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 ai-pa-browser-clawdbot
安装完成后，直接呼叫该 Skill 的名称或使用 /ai-pa-browser-clawdbot 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 ai-pa-browser-clawdbot skill. - Provides headless browser automation CLI for AI agents using accessibility tree snapshots and ref-based element selection. - Features support for multi-step web workflows, deterministic element selection, parallel browser sessions, and video recording. - Includes commands for navigation, element interaction, page snapshotting, state management, and advanced waiting. - Installation instructions and usage examples included in documentation.

元数据

Slug ai-pa-browser-clawdbot

版本 1.0.0

许可证 MIT-0

累计安装 1

当前安装数 1

历史版本数 1

常见问题