Skip to content
Get packs

Guide · Agent skill

How to use the BrowserWing Admin agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The browserwing-admin skill is a SKILL.md instruction file that teaches an AI agent how to operate BrowserWing, a browser automation platform that runs on your own machine. It is meant for people who already have BrowserWing and want an agent to manage its scripts, configure an LLM, and run automations. After following this guide you will be able to install the skill, trigger it with plain requests, and understand the API calls it makes.

BrowserWing AdminMemTensor/MemOS · Apache-2.0

Quick start

  1. 1Install BrowserWing and Chrome. Make sure BrowserWing is running on localhost:8080 and that Google Chrome is installed on the host. Verify Chrome with google-chrome --version or the macOS path given in the skill.
  2. 2Download and unzip the skill. Download the ZIP from this page and unzip it so SKILL.md sits in a browserwing-admin folder.
  3. 3Place the folder. Copy the folder into ~/.claude/skills/ for personal use or the project's .claude/skills/ folder, or upload the ZIP under Settings → Capabilities → Skills in the Claude apps.
  4. 4Add an LLM config. Ask the agent to add your LLM key, such as OpenAI with gpt-4o, and to test it. AI features need this step.
  5. 5Make a request. Describe the task in plain language, for example listing your scripts or running one with a keyword variable.
  6. 6Review and run. Read the curl commands the agent proposes, check the target site allows automation, then approve the run.

What is an agent skill and what does this one do?

An agent skill is a folder containing a SKILL.md file. The file gives an AI agent instructions for a specific task, and the agent reads it when a request matches. This skill documents BrowserWing's local HTTP API, so the agent knows which curl commands to run.

Once installed, the agent can help you install Google Chrome as a prerequisite, add and test an LLM API configuration, start an AI exploration that browses a website and produces a replayable script, create, update, run, or delete scripts, and export scripts as SKILL.md files for other agents.

  • Set up Google Chrome on Linux, macOS or Windows, or point BrowserWing at a remote Chrome with debugging enabled
  • Add, list, test, update and delete LLM configs for OpenAI, Anthropic, DeepSeek or any OpenAI-compatible endpoint
  • Run AI exploration and stream its events, then save the generated script
  • Create, list, play and inspect scripts, and review execution history
  • Export one or more scripts, or the executor browser API, as a SKILL.md file

Install the browserwing-admin skill

Download the ZIP from the skill's page on this directory and unzip it. The skill is a single file, SKILL.md. For a personal setup, place the unzipped folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.

In the Claude apps you can instead upload the ZIP under Settings → Capabilities → Skills.

  • Personal install: ~/.claude/skills/ (folder containing SKILL.md)
  • Project install: .claude/skills/ inside your project
  • Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Prepare BrowserWing and its requirements

The skill assumes BrowserWing is already running locally on port 8080. It does not install the platform itself. You also need Google Chrome on the host, and an LLM API key for AI features. A BrowserWing API key is optional and is used for authentication through the X-BrowserWing-Key header or an Authorization: Bearer token.

To check Chrome after installing it, run the version command shown in the skill. The skill lists install commands for Debian/Ubuntu and macOS (Homebrew cask) and a download link for Windows.

  • BrowserWing running on localhost:8080, or a remote Chrome with debugging enabled
  • Google Chrome installed on the host
  • An LLM API key such as OpenAI, Anthropic or DeepSeek (for AI features only)
  • Optional BrowserWing API key
  • curl or another HTTP client
google-chrome --version

Trigger the skill with natural requests

You do not call the skill by name. Describe the outcome you want in plain language and the agent will match the request to the skill's description, which covers installing dependencies, configuring the LLM, managing and executing scripts, AI exploration, the marketplace and troubleshooting.

Name the thing you want done and, where relevant, the site, the keyword or the script. The more specific the request, the fewer questions the agent needs to ask.

  • "Add my OpenAI key to BrowserWing as the default LLM config using gpt-4o."
  • "Use BrowserWing to explore bilibili.com, search for AI and save the script it makes."
  • "List my BrowserWing scripts and run the one that searches Bilibili with keyword deepseek."
  • "Export all my BrowserWing scripts to a SKILL.md file."

Run and export scripts

Scripts take variables written as ${variable_name} in action values. You pass values for them when you play a script. After a run, the agent can fetch extracted data from the last execution, for example content pulled by execute_js actions.

To share scripts with other agents, the skill exports them as a SKILL.md through POST /scripts/export/skill. Pass specific script IDs, or an empty array to export every script into one file. The exported file is then placed in another agent's skill directory so that agent can call POST /scripts/<id>/play.

  • Play a script by ID with a variables object
  • Read the latest extracted data from the play result endpoint
  • Review past runs with the script-executions history endpoint
  • Export selected scripts or all scripts into one SKILL.md
curl -X POST 'http://localhost:8080/api/v1/scripts/<script-id>/play' -H 'Content-Type: application/json' -d '{"variables": {"keyword": "deepseek"}}'

Safety notes before you run anything

This skill sends LLM API keys over the network and can control a real browser. That browser may be logged into your accounts, so run it deliberately, and avoid using it on shared or production machines without care. The AI explorer acts on live websites, so check that the site's terms allow automated access before exploring or replaying scripts against it.

The skill's metadata says it runs no scripts itself, but it uses the network. Its commands call a local API at localhost:8080 and, if you configure one, a remote Chrome control URL. Review each command before the agent runs it, especially those that delete scripts or change LLM configs.

  • Keep LLM keys out of shared logs and chat transcripts where possible
  • Use a separate browser profile or machine for automation that touches sensitive accounts
  • Confirm the target site allows automated access
  • The script marketplace section of the skill is marked as under development and its endpoint is not yet written out

What people use it for

  • Searching a site such as Bilibili for a keyword and collecting the first page of results on a schedule
  • Letting an AI explorer work out the steps for a repetitive web task, then saving the result as a replayable script
  • Pulling data from pages using execute_js actions and reading it back from the play result endpoint
  • Exporting a set of tested scripts as one SKILL.md so another agent can call them

Frequently asked questions

Is the browserwing-admin skill free to use?

The skill is published under the Apache-2.0 licence. BrowserWing itself must already be running on your machine, and AI features need an LLM API key from a provider such as OpenAI, Anthropic or DeepSeek, which you pay for separately.

What do I need before installing this skill?

You need BrowserWing running locally on port 8080 or a remote Chrome with debugging enabled, Google Chrome on the host, and curl or another HTTP client. An LLM API key is needed only for AI features, and a BrowserWing API key is optional.

Where do I put the SKILL.md file?

Put the unzipped browserwing-admin folder in ~/.claude/skills/ for personal use or in a project's .claude/skills/ folder. In the Claude apps, upload the ZIP under Settings → Capabilities → Skills instead.

Does this skill install BrowserWing?

No. The skill only documents BrowserWing's API and how to set up Chrome and LLM configs. You must already have BrowserWing installed and running.

Can the agent run scripts without asking me?

The skill uses the network and can control a real browser, so it is best to review each command before it runs. This matters most for deleting scripts, changing LLM configs, or exploring live sites.

How does this compare with writing my own browser automation?

This skill is a guide for operating BrowserWing through its local API, not a separate automation tool. If you do not want to run BrowserWing, or you need a different tool, you would need to look at other options, since this skill does not cover them.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

This skill suits developers and operations staff who already run BrowserWing and want an agent to create, run and export its scripts through ready-made curl commands. It is not useful unless BrowserWing is installed first, and it is not suited to shared or production machines or to sites whose terms forbid automation.

Download the skill