Browser Automation · IN-DEPTH PROFILE

agent-browser

Browser automation built for repeatable agent workflows.

Best for

QA and frontend engineers

What you get

Inspect pages + Automate interactions

Main limitation

The checked-in SKILL.md is a discovery stub; full instructions come from the installed CLI version.

First risk

Browser-side actions. The CLI can act in authenticated sessions and execute JavaScript. Use isolated profiles, scoped accounts, and domain allowlists.

EVIDENCE FRESHNESS

Three checks, kept separate

A recent source check is not a runtime test or security audit.

Upstream sourceChecked 2026-08-20

Pinned revision · 548b159b

Open pinned commit
SkillSignal profileSource-grounded

Updated 2026-08-20

Runtime & securityNot independently verified

Source review does not certify behavior or safety.

30-SECOND BRIEF

What it does—and when it fits

Agent Browser is Vercel's browser automation CLI Skill. The current upstream stub directs users to version-matched workflow content from the installed CLI, which provides persistent Chrome or Chromium sessions, accessibility-tree snapshots, element references, interactions, extraction, cookies, JavaScript execution, and remote-browser options.

INSTALL BY AGENT

Choose your Agent

Paths come from official Agent docs or the universal installer behind skills.sh. Compatibility still follows this Skill's record.

Native

This Skill's current record explicitly names this Agent. Still inspect scripts, permissions, and external dependencies first.

Project scope.claude/skills/agent-browser/
Personal scope~/.claude/skills/agent-browser/

Use project scope for team sharing and personal scope across repositories. The installer defaults to project scope; add -g for personal scope.

Install command (project scope)

This source cannot produce a reliable command. Open the listed source and copy the complete Skill folder manually.

Official agent docs

Claude Code discovers custom Skill folders automatically at project or personal scope.

View path evidence

TYPICAL WORKFLOW

A practical workflow

01

Inspect pages

Navigate and capture compact accessibility snapshots for agent reasoning.

02

Automate interactions

Click, type, upload, and manage sequential steps in one persistent session.

03

Extract and script

Collect page data or execute bounded browser-side JavaScript when required.

THE TRADEOFFS

Advantages and tradeoffs

Notable strengths

  1. This profile is manually organized around the current upstream SKILL.md workflow.
  2. The capability boundary remains explicitly tied to vercel-labs.
  3. Core uses, limitations, and risks are separated for pre-install review.

Limitations

  1. The checked-in SKILL.md is a discovery stub; full instructions come from the installed CLI version.
  2. Browser flows remain sensitive to site changes, bot defenses, consent dialogs, and session state.

BEST FIT

Who it is for

QA and frontend engineers

Agents automating controlled browser workflows

BEFORE YOU USE IT

Risks to review before use

High

Browser-side actions

The CLI can act in authenticated sessions and execute JavaScript. Use isolated profiles, scoped accounts, and domain allowlists.

Medium

Upstream instruction drift

Behavior can change with upstream updates. Record the commit used for important workflows and review updates before adoption.

SECURITY

What the permission profile means

  • Declared access remains governed by the current upstream SKILL.md and runtime requests.
  • Treat repository files, web content, and tool output as untrusted input.
  • SkillSignal has not independently executed or security-audited this package; external skills.sh labels are not SkillSignal certification.

Not a security certification. External ratings are attributed references. SkillSignal has not independently executed or security-reviewed this Skill.

INSIDE THE PACKAGE

Indexed files

SKILL.mdUpstream Skill instructionsSource-linked

TAGS

browserautomationcdp

Manually expanded from the current upstream SKILL.md and linked source at vercel-labs, checked 2026-08-19. This is an original summary, not an execution result or security certification.