BrowserAct Skills

BrowserAct Skills is the Agent Skill entry point for the browser-act CLI. The entry skill is intentionally small: it tells the agent when BrowserAct should be used, declares the CLI install path, and requires the agent to load current runtime guidance through browser-act get-skills core --skill-version 2.0.2 before running automation commands.

Use it when an agent needs a real browser workflow with indexed interactions, session ownership, screenshots, network capture, local or managed browser modes, and human handoff. It is a better fit for browser operation than for static HTTP fetching, and it should stay behind explicit authorization and confirmation boundaries.

Knowledge Freshness

The upstream skill declares version 2.0.2 and uses a two-layer design: a stable installed SKILL.md plus environment-aware runtime content returned by the CLI. That means the current browser list, active sessions, API-key status, directives, and command details are not fully represented in the static skill file.

Always run the documented get-skills core command at the start of a session and do not truncate its output. Verify CLI and skill compatibility before running sensitive browser operations.

Retrieval Sources

This listing is grounded in:

• The upstream browser-act/skills README.
• The BrowserAct entry SKILL.md.
• BrowserAct installation, skills, command reference, and Skill Forge docs.
• Repository license and current GitHub metadata.

Core Workflow

Install the CLI:

uv tool install browser-act-cli --python 3.12

Install the skill from:

https://github.com/browser-act/skills/tree/main/browser-act

Then have the agent load runtime instructions:

browser-act get-skills core --skill-version 2.0.2

For full browser automation, the docs use the loop:

open -> state -> interact -> state -> close

For example, after selecting an approved browser and target URL, an agent calls state, reads indexed elements, then uses commands such as click <index>, input <index> <text>, select <index> <option>, screenshot, or network requests.

Capability Scope

Use Cases

• Let an agent inspect and interact with a JavaScript-rendered site through a real browser.
• Use a supervised local Chrome session for an account workflow after explicit approval.
• Capture screenshots or network requests for debugging a web workflow.
• Hand off a difficult browser step to a human, then let the agent continue.
• Generate a reusable site-specific skill with Skill Forge after a reviewed exploration pass.
• Run multiple approved browser sessions without mixing cookies, account state, or task ownership.

Production Rules

• Use BrowserAct only on authorized websites, accounts, and datasets.
• Run browser-act get-skills core --skill-version 2.0.2 before the first browser command in each session.
• Preserve confirmation gates for browser creation, profile import, form submission, uploads, proxy changes, CAPTCHA assistance, and remote handoff.
• Prefer read-only inspection, screenshots, and state review before write actions such as clicks, form inputs, cookie changes, or uploads.
• Close sessions after use and review logs, screenshots, network captures, and generated Skill Forge packages before sharing them.

Source Review

Verified on 2026-06-18:

• GitHub metadata reported browser-act/skills as an MIT-licensed Python repository with topics including ai-agents, automation, claude-code-skills, openclaw-skills, codex-skill, web-scraping, data-extraction, cursor, and openclaw.
• The README describes BrowserAct as a browser automation CLI built for AI agents, with local Chrome, stealth browser modes, isolated sessions, indexed interaction, remote handoff, and Skill Forge.
• browser-act/SKILL.md declares skill name browser-act, version 2.0.2, the uv tool install browser-act-cli --python 3.12 install command, BrowserAct homepage, Python/uv requirements, local browser-profile storage, CDP permissions, confirmation requirements, and the get-skills core startup command.
• docs/installation.md documents agent-based skill installation, manual CLI install, optional API-key authentication for stealth browsers, stealth extraction, managed proxies, and CAPTCHA assistance, plus Python 3.12+ and Chrome/Chromium requirements.
• docs/skills.md documents the two-layer architecture: installed entry skill plus runtime content returned by get-skills, including environment state, browser list, sessions, core commands, and dynamic directives.
• docs/commands.md documents browser interaction, navigation, state, click/input/select, screenshots, network capture, cookies, CAPTCHA, remote assist, browser management, authentication, proxy management, and get-skills commands.
• docs/skill-forge.md documents the optional Skill Forge extension for generating reusable, parameterized skill packages from reviewed browser explorations.

Safety and Privacy

BrowserAct is powerful enough to act inside real websites and accounts. Keep the scope explicit: which site, which account, which browser profile, what data may be read, and what operations require confirmation. Do not let an agent import a local profile, submit a form, upload a file, buy proxy capacity, solve a challenge, or share a remote-assist URL without user approval.

The BrowserAct skill itself says sensitive data remains local except for CAPTCHA challenge images when solve-captcha is used. That is still a real data-flow decision. Avoid those features on sensitive pages, and review logs, screenshots, network captures, and generated Skill Forge packages before sharing them.

Duplicate Check

Checked current content/skills/, content/tools/, content/mcp/, content/agents/, guides, README entries, open pull requests, and repository-wide content for BrowserAct, browser-act/skills, browser-act skill, BrowserAct Claude Code skill, BrowserAct Codex skill, BrowserAct OpenClaw skill, Skill Forge, and matching source URLs. No dedicated BrowserAct Skills entry, exact source URL duplicate, target file, or open duplicate PR was found.

Disclosure

Editorial listing. No paid placement or affiliate link is used. BrowserAct Skills is MIT-licensed open-source content; BrowserAct's optional managed proxies, stealth browsers beyond the free tier, API-key features, CAPTCHA assistance, Discord support, target websites, browser providers, and generated skills may have separate licenses, pricing, privacy controls, and operational requirements.