SkillFoxx

BrowserSkill skill

BrowserSkill connects agents to signed-in Chrome or Edge and provides reading, clicks and extraction in a separate task window.

  • 10,344 ★ on the repository
  • Last commit
  • MIT
Skill

Broad access

Why this level

  • The agent acts with existing browser authentication and can submit forms or modify account data.
All reasons and checks

tencent/browserskillAuthor:tencent

Install

Manual install

Text for your agent

Set up browser-skill on this machine by following https://raw.githubusercontent.com/Tencent/BrowserSkill/main/AGENT_INSTALL.md

First Quick Start route: the prompt under Let your agent set it up. Read AGENT_INSTALL.md before execution; the user installs the extension. The manual CLI block follows.

Access and checks

How we review

Broad accessThe tool can change data, run commands or handle secrets.

  • The agent acts with existing browser authentication and can submit forms or modify account data.

This is third-party code. Review the repository files before installing.

What it does

Users can intervene in a browser task, after which the agent rereads page state. BrowserSkill connects a client to Chrome or Edge through `bsk`, a daemon and an extension. While the agent works, its actions remain visible in a separate window. It borrows an ordinary tab only with the user’s permission. The agent uses the selected browser profile’s saved authentication. To investigate a failure, it can read network responses and console messages or take screenshots. These records and images may expose internal site data, so the user decides which evidence to retain.

Who it is for. Browser test engineers working with an agent in their own signed-in profile and taking over the task when needed.

Good fit when

  • An internal page is already signed in and the agent reads its contents.
  • A user watches a form test in a separate window and intervenes when needed.
  • A loaded table is ready for JSON or CSV export with known row coverage.
  • A site error is reproducible and the agent correlates page state with requests and console output.

Not a fit when

  • The agent may not act in authenticated accounts.
  • Normal tabs would be borrowed without explicit permission.
  • Page content would be treated as new authority or instructions.

Example request

Read a test page in a separate BrowserSkill window, retain titles and links only and stop the session without submitting forms.

Limitations

Local operation requires the CLI, a running daemon, the extension and the skill in the chosen client. The extension supports Chrome and Edge on Chromium 125 or later; other browsers are not guaranteed to work. The agent acts through the selected profile’s signed-in accounts, so agree on the profile and data submission before work. Network responses, task recordings and screenshots may contain secrets; retain only the evidence needed for the task.

How to disable. Run `bsk session stop <id>` with the session identifier. Normal shutdown returns borrowed tabs to their original window and cleans up task pages. Disconnect the extension if needed.

FAQ

How do I start a session and obtain its ID?

Run `bsk session start --json` and retain the returned `session_id`. It associates subsequent commands with this particular task.

How do I check the connection?

Run `bsk doctor` and resolve any reported connection failures. The extension should show Connected.

Does it use my website authentication?

Yes. The agent uses the selected profile’s saved authentication and can act as the signed-in user. Agree on the destination and data before submitting a form.

How do I lend an ordinary tab to a task?

Have the agent list user tabs for the session. After user permission, it runs `bsk tab borrow <tab-id> --session <id>` with the selected tab and session identifiers. The tab moves into the task window.

Doctor passes but the skill is missing?

Install for the intended client with `bsk install-skill` and start a new session. Doctor does not load agent instructions.

Discussion

No questions or reviews yet.

Skills

A skill connects an agent to your Chrome through an extension and CLI for reading pages, managing tabs and verifying actions.

SkillBroad access649
Editors’ pick

A separate browser for agents with a skill and MCP: Claude Code and Codex work in your logged-in accounts, and every session is recorded for replay

SkillBroad access14.3K

A skill and CLI from Browser Use: the agent drives your real Chrome over the DevTools Protocol, logged in as you, and writes any missing helper itself

SkillBroad access19K
Editors’ pick

CLI, skill and Python library for browser control: the agent clicks, fills forms and reads pages over CDP

CLIBroad access120.8K

Web Access

web-access

A web access skill that picks between search, page fetch and driving your own browser over CDP

SkillBroad access9.4K

faster-chrome-devtools-skill drives Chrome over the DevTools Protocol from a shell, with no MCP server, Puppeteer or Playwright.

SkillBroad access173

README in short

Observations can be saved and reread. Moving a controlled tab out of its task window ends the session while preserving tabs.

README badge

Original SKILL.md

---
name: browser-skill
description: |
  Automate the user's logged-in Chromium browser: read pages, fill forms,
  scrape data, operate tabs, test a UI, or debug a website.
  Requires the bsk CLI and browser extension.
---

# browser-skill

Use `bsk` in an **Agent Window** with the user's existing logins. User tabs
require explicit borrowing. This skill does not install the extension or handle
advice-only tasks. Never extract credentials, cookies, tokens, or other secrets.

## Before acting

Human collaboration is automatic. Read [cooperation](references/cooperation.md) before acting; no enable/share call is needed.

- For website failures, request/performance investigations or reproduction evidence,
  read [debugging](references/debugging.md). Start capture before navigation or
  reproduction; ordinary browsing needs no capture.
- If a browser profile is required, read [tabs and profiles](references/tabs-and-profiles.md)
  before starting. Verify its instance mapping, bind every new session explicitly,
  and never substitute another instance or omit the selector to recover.
- Parallel work: [parallel tasks](references/tabs-and-profiles.md).
- In WorkBuddy/CodeBuddy or hosts that reap children, read [environment](references/environment.md)
  before sessions: reuse a reachable daemon or run `--foreground` in a managed background task.
  Read it also for missing CLI, startup/connection failures or remote pairing.
  Skill installation excludes the CLI/extension. Never restart shared daemons or
  delete runtime files to recover.

Report a problem