agentleFS
elxecutorelxecutor/skills/browser-use/SKILL.md

browser-use

Skill3 starsChanged yesterdayMITView on GitHub

Control the user's Chromium browser through CDP for interactive web tasks, authenticated pages, JavaScript-rendered sites, automation, testing, screenshots, and requested .onion browsing in Brave's Tor window. Prefer direct HTTP fetching for public static content unless the user requests browser interaction or Tor.

56 other projects use a version of this, 12 of them edited.

Adopted by

Most recently taken up by

Used with

Projects with it also use context7-efficient, configuring-dapr-pubsub and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • hfzarslanSkill

    context7-efficient

    Token-efficient library documentation fetcher using Context7 MCP with 86.8% token savings through intelligent shell pipeline filtering. Fetches code examples, API references, and best practices for JavaScript, Python, Go, Rust, and other libraries. Use when users ask about library documentation, need code examples, want API usage patterns, are learning a new framework, need syntax reference, or troubleshooting with library-specific information. Triggers include questions like "Show me React hooks", "How do I use Prisma", "What's the Next.js routing syntax", or any request for library/framework documentation.

    Also in 5 projects that use this

  • AbdullahMalik17Skill

    configuring-dapr-pubsub

    Configures Dapr pub/sub components for event-driven microservices with Kafka or Redis. Use when wiring agent-to-agent communication, setting up event subscriptions, or integrating Dapr sidecars. Covers component configuration, subscription patterns, publishing events, and Kubernetes deployment. NOT when using direct Kafka clients or non-Dapr messaging patterns.

    Also in 3 projects that use this

  • AbdullahMalik17Skill

    deploying-cloud-k8s

    Deploys applications to cloud Kubernetes (AKS/GKE/DOKS) with CI/CD pipelines. Use when deploying to production, setting up GitHub Actions, troubleshooting deployments. Covers build-time vs runtime vars, architecture matching, and battle-tested debugging.

    Also in 3 projects that use this

  • dycnnnbSkill

    remote-browser

    Controls a cloud browser from a sandboxed remote machine. Use when the agent is running in a sandbox (no GUI) and needs to navigate websites, interact with web pages, fill forms, take screenshots, or expose local dev servers via tunnels.

    Also in 3 projects that use this

  • mjunaidcaSkill

    building-chatgpt-apps

    Guides creation of ChatGPT Apps with interactive widgets using OpenAI Apps SDK and MCP servers. Use when building ChatGPT custom apps with visual UI components, embedded widgets, or rich interactive experiences. Covers widget architecture, MCP server setup with FastMCP, response metadata, and Developer Mode configuration. NOT when building standard MCP servers without widgets (use building-mcp-servers skill instead).

    Also in 3 projects that use this

  • mjunaidcaSkill

    building-chat-interfaces

    Build AI chat interfaces with custom backends, authentication, and context injection. Use when integrating chat UI with AI agents, adding auth to chat, injecting user/page context, or implementing httpOnly cookie proxies. Covers ChatKitServer, useChatKit, and MCP auth patterns. NOT when building simple chatbots without persistence or custom agent integration.

    Also in 3 projects that use this

  • +6 more from AbdullahMalik17

The file

We split this skill into the rules it gives an agent

6 rules. 0 are common, 1 is distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 54 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If the browser is not remote-debuggable, ask the user to enable it at chrome://inspect/#remote-debugging, then click Allow.

    line 55
  2. Distinctive:

    If Chrome is running but remote debugging is not enabled, open:

    line 59
  3. Very specific:

    Prefer to find elements with the accessibility tree, not screenshots: cdp("Accessibility.getFullAXTree")["nodes"] has every element's role, …

    line 69
  4. Very specific:

    After navigation, call wait_for_load().

    line 72
  5. Very specific:

    Use js(...) for DOM inspection or extraction when coordinates are the wrong tool.

    line 74
  6. Very specific:

    Use watch_network() to see network requests in a dev-tool-like table.

    line 75

“Common” is measured: kept by at least 6 in 10 of the 54 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Browser Use

Direct browser control via CDP. Use ~/.agents/skills/browser-use/bu. It installs its own namespaced Linux cache via UV when absent and controls the user's default Chromium browser through its approved remote-debugging connection. The cache is $XDG_CACHE_HOME/opencode-skills/browser-use, not a shared global tool install.

When Not to Use

A basic fetch of public information needs no browser. If a plain HTTP request can read it — a public page, an API, docs — use curl or your fetch tool, and leave the browser alone. Use browser-use when the task needs interaction (click, type, navigate), the user's logged-in session, JS rendering, or a bot-protected page. If a direct fetch fails or returns a shell page, then escalate to the browser.

When the user requests a visible browser tab or browsing over Tor, use the requested browser/session instead of substituting an HTTP probe. For .onion sites in Brave's native Tor window, follow the onion/Tor workflow: identify the user-confirmed Tor tab and preserve its context. Do not switch to a normal browser, proxy, or different mirror without permission.

Usage

~/.agents/skills/browser-use/bu <<'PY'
print(page_info())
PY
  • Invoke the skill-local bu wrapper. Use heredocs for multi-line commands.
  • Helpers are pre-imported.
  • For ordinary browsing, first navigation is new_tab(url), not goto_url(url). This default does not select an existing Brave Tor context; use the onion/Tor workflow instead for that case.

Local Chrome

The first real bu invocation starts/reuses Browser Harness' local daemon and keeps its CDP connection open while Chrome displays any Allow prompt. Wait for the user to click Allow; do not give up, open another connection, or run probes. --doctor only reports health; it does not start the daemon. After a definite daemon crash, reload it once and retry the original command:

~/.agents/skills/browser-use/bu <<'PY'
ensure_real_tab()
print(page_info())
PY

# Only after a daemon crash, not while an Allow prompt is pending:
~/.agents/skills/browser-use/bu --reload

Diagnostics:

~/.agents/skills/browser-use/bu --doctor

If the browser is not remote-debuggable, ask the user to enable it at chrome://inspect/#remote-debugging, then click Allow. An explicit BU_CDP_WS can intentionally select a remote browser endpoint.

If Chrome is running but remote debugging is not enabled, open:

chrome://inspect/#remote-debugging

Tick "Allow remote debugging for this browser instance" and click Allow if a permission popup appears. Then retry.

Page Workflow

  • Prefer to find elements with the accessibility tree, not screenshots: cdp("Accessibility.getFullAXTree")["nodes"] has every element's role, name, and backendDOMNodeId — filter in Python before printing (it is thousands of nodes). Coordinates: q = cdp("DOM.getBoxModel", backendNodeId=n)["model"]["content"]; x, y = sum(q[0::2])/4, sum(q[1::2])/4 (viewport px, ready for click_at_xy; negative/oversized means scroll first).
  • Clicking: AX node -> box center -> click_at_xy(x, y) -> verify with a targeted js(...)/page_info() check.
  • Fall back to raw HTML via js(...) only when the AX tree lacks the element (canvas, exotic widgets); screenshot when layout or imagery matters.
  • After navigation, call wait_for_load().
  • For ordinary browsing, if the current tab is stale or internal, call ensure_real_tab(). Do not use it to pick a Brave Tor tab: it may select a normal tab, while a new Tor tab may itself be an internal page.
  • Use js(...) for DOM inspection or extraction when coordinates are the wrong tool.
  • Use watch_network() to see network requests in a dev-tool-like table.
  • Login walls: stop and ask. Exception: use available SSO automatically when Chrome is already signed in; still stop for passwords, MFA, consent, or ambiguous account choice.
  • Raw CDP is available with cdp("Domain.method", ...).

Recordings

~/.agents/skills/browser-use/bu recordings enable
~/.agents/skills/browser-use/bu recordings disable
~/.agents/skills/browser-use/bu recordings

Gotchas

  • chrome://inspect/#remote-debugging must be enabled for local Chrome control.
  • Omnibox popups are not real work tabs.
  • CDP target order is not Chrome's visible tab-strip order.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.