google-ui-automation
Skill0 starsChanged 48 days agoMITView on GitHub
Field notes for driving Google Ads, GA4, Search Console and Tag Manager web UIs through browser automation (Claude in Chrome or similar) — frozen renderers, deep links that 404, multi-login accounts, typing into Material inputs, tab-group quirks, and which steps must be left to the human. Read before automating any Google marketing UI.
In this repository7 other files it gives its agents.
- ad-assetsSkill
- conversion-trackingSkill
- google-growth-journeySkill
- growth-auditSkill
- pmax-campaignSkill
- search-console-seoSkill
- tag-managerSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 12
Prefer reading text (get_page_text, innerText via JS) over screenshots on Ads pages.
- Distinctive: line 24
If a tab freezes: open a new tab, don't keep retrying the frozen one.
- Distinctive: line 43
Delete/remove campaigns or links without an explicit request (removal is permanent).
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Automating Google's marketing UIs
These UIs are heavy Angular/Material apps. Most of the time lost is not in the task but in fighting the page. Rules learned the hard way:
Reading the page
- Prefer reading text (
get_page_text,innerTextvia JS) over screenshots on Ads pages. Screenshots on heavy Ads views time out ("renderer may be frozen") while text extraction still works. - A summary/overview screen may show stale or partial data (blank campaign name, default asset-group name, "all countries"). Open the actual settings step to confirm.
- Some pages block reading
location.href/ query strings through the automation tool; read the rendered text instead.
Navigating
- Many old deep links 404 (
/aw/tagmanagement,/aw/linkedaccounts). Navigate from the menu: Tools → Data manager [Trình quản lý dữ liệu] is where links and the Google tag live. - GA4 hash routes (
#/a…p…/admin/...) sometimes bounce to Home; go to Admin and click. - If a tab freezes: open a new tab, don't keep retrying the frozen one. If another connected browser has the same login, it may render the page fine.
- Multiple Google logins: the
authuser=/ocid=/__u=parameters bind a URL to one login. An account only reachable from another login redirects to the default account — check the account number in the header after every navigation.
Typing and clicking
- Material inputs often ignore the first typed text right after page load. Click, wait ~1 s, type, then zoom/read the field to confirm before pressing Enter.
Ctrl+Ain some fields types "a" instead of selecting; clear with the field's ✕ button.- Batch-typing into many fields can insert stray characters; re-read every field after.
- Toast/confirmation overlays ("Indexing requested", "Get to know…") block clicks underneath; dismiss them first.
- Buttons that don't respond to a coordinate click: find the element by text and
click()it via JS, then re-read the page.
What the agent must not do
- Complete identity verification, enter passwords/2FA, add payment methods or top up.
- Press Publish on a campaign unless the user explicitly says so for that campaign.
- Delete/remove campaigns or links without an explicit request (removal is permanent).
- Accept terms or consent on the user's behalf. Prepare everything up to those buttons and report exactly what is waiting for the human.
Verification helpers
- YouTube video owner:
https://www.youtube.com/oembed?url=https://youtu.be/<id>&format=json. - Live tag state:
${CLAUDE_PLUGIN_ROOT}/scripts/check_gtag.jsin the page console.
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
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.

