skills.rest
dingdyan
dingdyanCOMMUNITY
Explore all skills by this author →
GitHub

Agent Browser

Automate web navigation and UI interactions using accessibility-tree snapshots.

Updated Apr 5, 2026
Search Tags:#debugging#UI testing#form filling#web automation#session state#headless browser#element snapshots
One-click install
npx skills add https://github.com/dingdyan/openclaw-workspace-v2 --skill agent-browser-dingdyan
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: Agent Browser
Source: https://github.com/dingdyan/openclaw-workspace-v2/tree/main/skills/agent-browser
Command: npx skills add https://github.com/dingdyan/openclaw-workspace-v2 --skill agent-browser-dingdyan

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It solves the problem of manually navigating websites and performing repetitive UI interactions when you need reliable, step-by-step browser automation for AI agents.

Core Features & Use Cases

  • Structured web navigation and interaction: Open pages, click, fill, type, and run common control actions while keeping steps explicit and repeatable.
  • Snapshot-driven element targeting: Use accessibility-tree snapshots to reference elements (for example, click or fill by @ref) and re-snapshot after DOM changes.
  • Evidence output for verification: Support JSON outputs, screenshots, and recordings to make results easier to debug and validate.
  • Use Case: Automate a multi-step form workflow on a web app by opening the page, snapshotting interactive elements, filling fields by ref, waiting for the expected state, and verifying the outcome with a new snapshot or screenshot.

Quick Start

Run agent-browser open https://example.com and then agent-browser snapshot -i to obtain interactive element refs you can use for subsequent click and fill steps.

Frequently Asked Questions about Agent Browser

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate web navigation and form filling for AI agents?▼

Automate web navigation and form filling by issuing deterministic command sequences to open pages, snapshot interactive elements by accessibility-tree refs, and fill fields explicitly. You re-snapshot after DOM changes and verify outcomes using screenshots or JSON outputs.

How does snapshot-driven element targeting work for UI testing automation?▼

Snapshot-driven element targeting works by capturing accessibility-tree snapshots to reference interactive elements via @ref. You use these refs to execute deterministic click or fill actions, ensuring repeatable UI testing automation across dynamic websites.

What's the best way to debug dynamic web page interactions during browser automation?▼

The best way to debug dynamic web page interactions is to produce machine-readable JSON outputs, screenshots, and recordings during browser automation. These evidence outputs make it easier to validate whether your deterministic command sequencing achieved the expected page or network state.

Can I use a headless browser to execute deterministic UI interactions and wait for network conditions?▼

Yes, you can use a headless browser to execute deterministic UI interactions. The automation supports controlled browser settings, session state management, waiting for specific page or network conditions, and producing machine-readable outputs for verification.

Do I need to re-snapshot interactive elements after DOM changes when automating web tasks?▼

Yes, you need to re-snapshot interactive elements after DOM changes when automating web tasks. Snapshot-driven element targeting requires capturing new accessibility-tree snapshots to ensure your @ref pointers remain valid for subsequent click and fill steps.

Why are my automated browser commands failing to interact with dynamic website elements?▼

Automated browser commands fail to interact with dynamic website elements when accessibility-tree snapshot refs become stale after DOM updates. You must re-snapshot interactive elements after page state changes and wait for expected network conditions before executing subsequent deterministic commands.

Related Skills

Explore More Software Engineering Skills→
swongsuddeeswongsuddee
1

playwright-cli

Automate browser interactions and capture evidence using Playwright CLI.

Community
Intermediate
svosvo

verification-before-completion

Require fresh verification outputs before allowing completion claims.

Community
Advanced
joneqianjoneqian

agent-browser

Control a headless browser using accessibility-tree snapshots and ref-based element targeting.

Community
Advanced
RonkTsangRonkTsang
20

agent-browser

Automates web interactions via Bash scripts for navigation, form filling, and data extraction.

Community
Advanced