BumpyClock/agent-templates / skills_archive/browser-use
browser-use
VerifiedChrome DevTools MCP automation: existing Chrome tabs, no AppleScript.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 3.6 KB |
| agents/openai.yaml | config | 218 B |
| attribution.txt | markdown | 168 B |
| mcporter-config.md | markdown | 3.9 KB |
SKILL.md
View on GitHub---
name: browser-use
description: "Chrome DevTools MCP automation: existing Chrome tabs, no AppleScript."
---
# Browser Use
Browser tasks against existing Chrome session. Config repair: `mcporter-config.md`.
Hard rule: reattach existing Chrome profile only.
```bash
mcporter call chrome-devtools.<tool>
```
Login-heavy sites fail in isolated profiles (captcha, device checks, missing SSO/extension). Prefer existing Chrome for any site needing login.
Never use `chrome-isolated`, Playwright, Puppeteer, Codex in-app browser, AppleScript, `osascript`, GUI scripting, or macOS `open` for browser control unless user explicitly asks for isolated/new browser.
Screenshot/live UI bugs require existing-Chrome path. `curl`, source inspection, Worker tests, local Playwright = supporting proof only; not equivalent when user showed rendered browser problem or page depends on login/profile state.
## Check MCP
```bash
mcporter list chrome-devtools --schema
mcporter call chrome-devtools.list_pages --args '{}' --output text
```
`list_pages` must show user's real open tabs. Blank/default isolated Chrome → reattach failed, stop and say so.
Attach prompt recovery: if Chrome shows "Allow remote debugging?", click Allow via Peekaboo, rerun `list_pages`.
```bash
PB="${PEEKABOO_BIN:-$HOME/bin/peekaboo}"
[ -x "$PB" ] || PB="$(command -v peekaboo)"
"$PB" permissions status --json
"$PB" see --app frontmost --path /tmp/chrome-attach.png --json --annotate
# Click visible Allow button using current snapshot coords
"$PB" click --coords <allow_x>,<allow_y> --json
mcporter call chrome-devtools.list_pages --args '{}' --output text
```
Attach-alert rule: snapshot shows Chrome DevTools/MCP attach prompt → click visible Allow once, rerun `list_pages`. Button not visible or prompt ambiguous → stop, ask. Never silently switch to Playwright/Puppeteer.
`DevToolsActivePort` error → ask user to restart Chrome or DevTools bridge, retry once:
```bash
mcporter daemon restart
mcporter call chrome-devtools.list_pages --args '{}' --output text
```
Still fails → stop, say Chrome DevTools MCP unavailable. No AppleScript.
Avoid noisy recovery loops. Repeated restarts trigger reconnect/login prompts. Try once, then pause.
## Typical Flow
```bash
mcporter call chrome-devtools.select_page --args '{"pageId":9}' --output text
mcporter call chrome-devtools.take_snapshot --args '{}' --output text
mcporter call chrome-devtools.navigate_page --args '{"url":"https://example.com"}' --output text
mcporter call chrome-devtools.click --args '{"uid":"1_38","includeSnapshot":true}' --output text
mcporter call chrome-devtools.fill --args '{"uid":"1_13","value":"text","includeSnapshot":true}' --output text
mcporter call chrome-devtools.evaluate_script --args '{"function":"() => document.title"}' --output json
```
Use `take_snapshot` before actions, current `uid` values only. `take_screenshot` only when visual layout matters.
## Live UI Proof
```bash
mcporter call chrome-devtools.list_pages --args '{}' --output text
mcporter call chrome-devtools.select_page --args '{"pageId":9}' --output text
mcporter call chrome-devtools.navigate_page --args '{"url":"https://example.com"}' --output text
mcporter call chrome-devtools.take_snapshot --args '{}' --output text
mcporter call chrome-devtools.evaluate_script --args '{"function":"() => document.body.innerText"}' --output json
```
Browser automation unavailable → report as verification gap, don't substitute isolated tooling.
## Secret Handling
Never print tokens/passwords from DOM, network logs, inputs. Token checks: shape only (present/absent, length, status code, account/org).
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 470k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |