Agent Browser Clawdbot

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

Security Scan

The skill's instructions, scope, and requirements are coherent with a headless browser automation CLI; no code or install artifacts are included in the package, but verify the external CLI before installing and be cautious with saved auth/state files.

Purpose & Capability

The name/description match the SKILL.md: it documents a CLI 'agent-browser' for deterministic, sessioned browser automation. Minor metadata inconsistencies exist: the provided _meta.json ownerId and version differ from the registry metadata (owner id and version mismatch), and the registry 'Source' is 'unknown' while SKILL.md points to github.com/vercel-labs/agent-browser—these are likely bookkeeping issues but worth verifying against the upstream project before trusting installs.

Instruction Scope

The instructions stay within the browser-automation domain: navigation, snapshots, ref-based interactions, sessions, state save/load, network routing and mocking. They do reference loading/saving state files (e.g., auth.json, admin-auth.json) and an optional AGENT_BROWSER_SESSION env var; these are expected for this tool but are sensitive operations (they read/write authentication/session data).

Install Mechanism

No install spec is included in the skill bundle (instruction-only). The SKILL.md suggests installing via npm and an 'agent-browser install' step to download Chromium—this is external to the skill and should be validated by the user (check the npm package name and upstream repo).

Credentials

The skill declares no required environment variables, which aligns with the bundle. The documentation mentions an optional AGENT_BROWSER_SESSION env var and uses state save/load files; although optional, these mechanisms can carry secrets (cookies/storage). No unrelated credentials or env vars are requested by the skill itself.

Persistence & Privilege

The skill is not always-enabled and does not request system persistence. It is instruction-only and won't write code to disk by itself. However, the external CLI (when installed by the user) will store state files if used—this is normal for a browser automation tool.

Assessment

This skill is an instruction-only wrapper that expects you to have the external 'agent-browser' CLI installed. Before installing or using it: 1) Verify the upstream project (github.com/vercel-labs/agent-browser) and the npm package name/owner to avoid typosquats or malicious forks. 2) Be cautious with state files (auth.json, admin-auth.json) — only load state files you trust because they can contain cookies and other credentials. 3) Network-mocking and request-interception features are normal for a browser automation tool but can be misused; restrict what agent tasks are allowed to browse and avoid giving the agent access to sensitive internal sites unless necessary. 4) Note the metadata mismatch in _meta.json vs registry (owner/version) — treat that as a sign to double-check provenance. If you need extra assurance, obtain the CLI directly from the official repository and inspect the package before installation.

Agent Browser Skill

Fast browser automation using accessibility tree snapshots with refs for deterministic element selection.

Why Use This Over Built-in Browser Tool

Use agent-browser when:

• Automating multi-step workflows
• Need deterministic element selection
• Performance is critical
• Working with complex SPAs
• Need session isolation

Use built-in browser tool when:

• Need screenshots/PDFs for analysis
• Visual inspection required
• Browser extension integration needed

Core Workflow

# 1. Navigate and snapshot
agent-browser open https://example.com
agent-browser snapshot -i --json

# 2. Parse refs from JSON, then interact
agent-browser click @e2
agent-browser fill @e3 "text"

# 3. Re-snapshot after page changes
agent-browser snapshot -i --json

Key Commands

Navigation

agent-browser open <url>
agent-browser back | forward | reload | close

Snapshot (Always use -i --json)

agent-browser snapshot -i --json          # Interactive elements, JSON output
agent-browser snapshot -i -c -d 5 --json  # + compact, depth limit
agent-browser snapshot -s "#main" -i      # Scope to selector

Interactions (Ref-based)

agent-browser click @e2
agent-browser fill @e3 "text"
agent-browser type @e3 "text"
agent-browser hover @e4
agent-browser check @e5 | uncheck @e5
agent-browser select @e6 "value"
agent-browser press "Enter"
agent-browser scroll down 500
agent-browser drag @e7 @e8

Get Information

agent-browser get text @e1 --json
agent-browser get html @e2 --json
agent-browser get value @e3 --json
agent-browser get attr @e4 "href" --json
agent-browser get title --json
agent-browser get url --json
agent-browser get count ".item" --json

Check State

agent-browser is visible @e2 --json
agent-browser is enabled @e3 --json
agent-browser is checked @e4 --json

Wait

agent-browser wait @e2                    # Wait for element
agent-browser wait 1000                   # Wait ms
agent-browser wait --text "Welcome"       # Wait for text
agent-browser wait --url "**/dashboard"   # Wait for URL
agent-browser wait --load networkidle     # Wait for network
agent-browser wait --fn "window.ready === true"

Sessions (Isolated Browsers)

agent-browser --session admin open site.com
agent-browser --session user open site.com
agent-browser session list
# Or via env: AGENT_BROWSER_SESSION=admin agent-browser ...

State Persistence

agent-browser state save auth.json        # Save cookies/storage
agent-browser state load auth.json        # Load (skip login)

Screenshots & PDFs

agent-browser screenshot page.png
agent-browser screenshot --full page.png
agent-browser pdf page.pdf

Network Control

agent-browser network route "**/ads/*" --abort           # Block
agent-browser network route "**/api/*" --body '{"x":1}'  # Mock
agent-browser network requests --filter api              # View

Cookies & Storage

agent-browser cookies                     # Get all
agent-browser cookies set name value
agent-browser storage local key           # Get localStorage
agent-browser storage local set key val

Tabs & Frames

agent-browser tab new https://example.com
agent-browser tab 2                       # Switch to tab
agent-browser frame @e5                   # Switch to iframe
agent-browser frame main                  # Back to main

Snapshot Output Format

{
  "success": true,
  "data": {
    "snapshot": "...",
    "refs": {
      "e1": {"role": "heading", "name": "Example Domain"},
      "e2": {"role": "button", "name": "Submit"},
      "e3": {"role": "textbox", "name": "Email"}
    }
  }
}

Best Practices

1. Always use -i flag - Focus on interactive elements
2. Always use --json - Easier to parse
3. Wait for stability - agent-browser wait --load networkidle
4. Save auth state - Skip login flows with state save/load
5. Use sessions - Isolate different browser contexts
6. Use --headed for debugging - See what's happening

Example: Search and Extract

agent-browser open https://www.google.com
agent-browser snapshot -i --json
# AI identifies search box @e1
agent-browser fill @e1 "AI agents"
agent-browser press Enter
agent-browser wait --load networkidle
agent-browser snapshot -i --json
# AI identifies result refs
agent-browser get text @e3 --json
agent-browser get attr @e4 "href" --json

Example: Multi-Session Testing

# Admin session
agent-browser --session admin open app.com
agent-browser --session admin state load admin-auth.json
agent-browser --session admin snapshot -i --json

# User session (simultaneous)
agent-browser --session user open app.com
agent-browser --session user state load user-auth.json
agent-browser --session user snapshot -i --json

Installation

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

Credits

Skill created by Yossi Elkrief (@MaTriXy)

agent-browser CLI by Vercel Labs