Agent skill · Testing & QA

agent-browser

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction.

majiayu000github.com/majiayu000GitHub ↗
claude-coderead-onlyMIT
Install
npx skills add majiayu000/claude-skill-registry --skill 3rd-agent-browser --agent claude-code

Same command for any agent — swap --agent for codex, cursor, copilot.

Facts
Files in the skill folder: 2
SKILL.md size: 18 KB
Bundled scripts: none
Allowed tools: Bash(npxagent-browser:*)Bash(agent-browser:*)
Path: skills/agent/3rd-agent-browser/SKILL.md
Open the folder on GitHub →
Where it comes from
Stars: 534
Language: HTML

Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.

Review
written from the skill's own SKILL.md · Aug 5, 2026

What it does

Automates browser tasks for AI agents, enabling navigation, element discovery, form filling, clicking, waiting, taking screenshots, and data extraction across web pages. It provides a structured workflow: open a URL, snapshot to obtain element refs, perform interactions using those refs, then re-snapshot after DOM updates. It includes commands for navigation, interaction, information retrieval, waiting, downloads, capture, diffing, session management, and various patterns (e.g., forms, auth, state, parallel sessions, and mobile/iOS). It also supports features like authentication with saved profiles, state persistence, and visual debugging via annotated screenshots. It documents a broad set of essential commands for interacting with web content.

How it works

The skill instructs the agent to:

  • Navigate with agent-browser open <url> and close when done with agent-browser close.
  • Snapshot interactive elements with agent-browser snapshot -i to obtain refs like @e1, @e2, @e3.
  • Perform interactions using those refs, including click, fill, type, select, check, and press.
  • Re-snapshot after navigation or DOM changes to refresh refs.
  • Chain commands with && for efficiency since a background daemon preserves the browser state.
  • Use various patterns for forms, authentication (auth save/login/show/delete), and session persistence.
  • Retrieve information with get text @e1, get url, and get title.
  • Wait for conditions with wait variants like --load networkidle, specific selectors, URLs, or timeouts.
  • Capture outputs with screenshot, pdf, or diff against baselines, including visual diffs.
  • Manage sessions and state across multiple agents and sessions, including named sessions and state files.
  • Apply optional features such as content boundaries, domain allowlists, action policies, and output limits.
  • Use advanced capabilities like annotated screenshots, semantic locators, and JavaScript evaluation via eval.

When to use it

Use when interacting with websites or web apps programmatically is required, including navigating pages, filling forms, clicking buttons, taking screenshots, scraping data, testing, logging in, or automating any browser task. Triggers include requests to open a website, fill out a form, click a button, take a screenshot, scrape data, test a web app, login, automate browser actions, or similar tasks requiring web interaction.

What it can touch

  • Commands and tool names exactly as described, including agent-browser with subcommands like open, snapshot, fill, click, wait, screenshot, pdf, diff, auth, state, session, etc.
  • It references inputs and outputs such as URLs, element refs like @e1, and file paths for downloads or screenshots.

Caveats

  • Security features are opt-in; default navigation and actions have no enforced restrictions unless configured.
  • Requires certain environments for mobile/iOS or local file access (macOS with Xcode and Appium for iOS simulator).
  • Uses environment variables for boundaries, domain allowlists, action policies, and max output limits.
  • Refs are invalidated when the page changes; a new snapshot is required after navigation or dynamic content loads.
  • Output and behavior depend on the browser automation runtime and its supported features, such as CDP connections or a background daemon.
From the SKILL.md

# Browser Automation with agent-browser ## Core Workflow Every browser automation follows this pattern: 1. **Navigate**: `agent-browser open <url>` 2. **Snapshot**: `agent-browser snapshot -i` (get element refs like `@e1`, `@e2`) 3. **Interact**: Use refs to click, fill, select 4. **Re-snapshot**: After navigation or DOM changes, get fresh refs ```bash agent-browser open https://example.com/form agent-browser snapshot -i # Output: @e1 [input type="email"], @e2 [input type="password"], @e3 [button] "Submit" agent-browser fill @e1 "user@example.com" agent-browser fill @e2 "password123" agent-browser click @e3 agent-browser wait --load networkidle agent-browser snapshot -i # Check result ``` ## Command Chaining Commands can be chained with `&&` in a single shell invocation. The browser persists between commands via a background daemon, so chaining is safe and more efficient than separate calls. ```bash # Chain open + wait + snapshot in one call agent-browser open https://example.com && agent-browser wait --load networkidle && agent-browser snapshot -i # Chain multiple interactions agent-browser fill @e1 "user@example.com" && agent-browser fill @e2 "password123" && agent-browser click

What's inside
Steps it walks through
  1. Core Workflow
  2. Command Chaining
  3. Essential Commands
  4. Common Patterns
  5. Form Submission
  6. Authentication with Auth Vault (Recommended)
  7. Authentication with State Persistence
  8. Session Persistence
  9. Data Extraction
  10. Parallel Sessions
  11. Connect to Existing Chrome
  12. Color Scheme (Dark Mode)
  13. Visual Browser (Debugging)
  14. Local Files (PDFs, HTML)
Ships with 1 file
  • metadata.json
Commands it runs
agent-browser open https://example.com/form
agent-browser snapshot -i
agent-browser fill @e1 "user@example.com"
agent-browser fill @e2 "password123"
agent-browser click @e3
agent-browser wait --load networkidle
agent-browser snapshot -i  # Check result
Chain open + wait + snapshot in one call
agent-browser open https://example.com && agent-browser wait --load networkidle && agent-browser snapshot -i
Chain multiple interactions
More from claude-skill-registry
All skills →
descriptive-statisticsSummarizing and visualizing data through measures of center, spread, shape, and position. Covers…Data & Analyticsperplexity-researcher-reasoning-proHighest level of research and reasoning capabilities for complex decision-making with significant…OtherretroBrutally honest engineering retrospective. Analyzes git history for what actually happened — not…Data & AnalyticsanalyzeUniversal multi-perspective analyzer for any topic, file, idea, or decision. Extract key points…Otherarize-prompt-optimizationINVOKE THIS SKILL when optimizing, improving, or debugging LLM prompts using production trace data…Data & Analyticscellxgene-contextGuide for integrating CellxGene Census single-cell data with ENCODE bulk experiments. Use when…Data & Analytics
About this skill
What does the agent-browser skill do?

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction.

How do I install it?

Run `npx skills add majiayu000/claude-skill-registry --skill 3rd-agent-browser --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.

Where does this skill come from?

From majiayu000/claude-skill-registry, a repository with 534 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.

Is a popular skill a good skill?

Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.

Keep going
All skills →
The full catalog, by category and by tool.
Open Source Radar →
The repositories these skills come from, watched daily.
The people behind them →
Makers and builders across the ecosystem.