browser-harness

name

browser-harness

description

Direct browser control via CDP — automate, scrape, test, or interact with web pages by driving the user's already-running Chrome (or a Browser Use cloud browser). Use when the user wants to click, screenshot, fill forms, extract data, or navigate real web pages. Default to screenshots + coordinate clicks, not selector hunting. Requires the one-time `browser-harness` CLI install (see references/install.md).

browser-harness

Direct browser control via CDP. You drive the user's real browser with Python helpers run through the browser-harness command.

Prerequisite (one-time — NOT part of the AI workflow)

This skill is instructions only. It assumes the browser-harness command is already on $PATH. If command -v browser-harness fails, do the one-time install in references/install.md first, then continue. Installation and browser-connection setup are a prerequisite; once browser-harness <<'PY' … PY prints page info, never run install/connection steps again as part of normal work.

Usage

browser-harness <<'PY'
new_tab("https://docs.browser-use.com")
wait_for_load()
print(page_info())
PY

• Invoke as browser-harness — it's on $PATH after install. No cd, no uv run.
• Use the heredoc form for every multi-line command. It prevents shell quote mangling inside Python strings and JavaScript snippets.
• First navigation is new_tab(url), not goto_url(url) — goto runs in the user's active tab and clobbers their work.
• Helpers are pre-imported and the daemon auto-starts; you never start/stop it manually unless you want to.

What actually works

• Screenshots first. capture_screenshot() to understand the page, find visible targets, and decide whether you need a click, a selector, or more navigation.
• Clicking. capture_screenshot() → read the pixel off the image → click_at_xy(x, y) → capture_screenshot() to verify. Suppress the Playwright-habit reflex of "locate first, then click" — no getBoundingClientRect, no selector hunt. Drop to DOM only when the target has no visible geometry. Hit-testing happens in Chrome's browser process, so clicks pass through iframes / shadow DOM / cross-origin without extra work.
• Bulk HTTP. http_get(url) + ThreadPoolExecutor. No browser needed for static pages.
• After goto: wait_for_load().
• Wrong/stale tab: ensure_real_tab().
• Verification: print(page_info()) is the simplest "is this alive?" check; screenshots are the default way to verify whether a visible action worked.
• DOM reads: use js(...) for inspection/extraction when a screenshot shows coordinates are the wrong tool.
• Auth wall: redirected to login → stop and ask the user. Don't type credentials from screenshots.
• Raw CDP for anything helpers don't cover: cdp("Domain.method", params).

After every meaningful action, re-screenshot before assuming it worked.

Remote / cloud browsers

Use remote for parallel sub-agents (each gets an isolated browser via a distinct BU_NAME) or on a headless server. BROWSER_USE_API_KEY must be set.

browser-harness <<'PY'
start_remote_daemon("work")   # clean cloud browser; profileName=/profileId= to reuse a logged-in profile
PY

BU_NAME=work browser-harness <<'PY'
new_tab("https://example.com")
print(page_info())
PY

start_remote_daemon prints a liveUrl so the user can watch. Running remote daemons bill until timeout.

Interaction skills (progressive disclosure)

If you struggle with a specific UI mechanic, read the matching file under ${CLAUDE_PLUGIN_ROOT}/interaction-skills/ before inventing an approach. Available: browser-wall, connection, cookies, cross-origin-iframes, dialogs, downloads, drag-and-drop, dropdowns, iframes, network-requests, print-as-pdf, profile-sync, screenshots, scrolling, shadow-dom, tabs, uploads, viewport.

Task-specific edits

For task-specific helper additions, edit ${CLAUDE_PLUGIN_ROOT}/agent-workspace/agent_helpers.py. Keep core helpers short.

Domain skills (opt-in)

Community per-site playbooks live in ${CLAUDE_PLUGIN_ROOT}/agent-workspace/domain-skills/<host>/ and are off by default. Set BH_DOMAIN_SKILLS=1 to enable them; when enabled and the task is site-specific, read every file in the matching <site>/ directory before inventing an approach.

Design constraints

• Coordinate clicks default. Input.dispatchMouseEvent goes through iframes/shadow/cross-origin at the compositor level.
• Connect to the user's running Chrome. Don't launch your own browser.
• Prefer compositor-level actions (screenshots, coordinate clicks, raw key input) over framework/DOM hacks. Reach for interaction-skills/ only when those are the wrong tool.