mobile-verification
Skill1 starsChanged 4 months agoMITView on GitHub
Test what a page actually DOES on a phone — real taps, real scrolling, real animations — using Playwright instead of a browser-automation tab that is structurally blind to motion. Use whenever verifying mobile behaviour, testing touch interactions, checking horizontal overflow or touch-target sizes, before merging any front-end change, or when a scroll/animation/click test reports "nothing happened" and you need to know whether the page is broken or your harness is.
- Installs packages
In this repository11 other files it gives its agents.
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are 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.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 34
Before you ever conclude "the page can't scroll":
- Distinctive: line 79
A harness that has never failed cannot be trusted: If it can't fail on a false assertion, its passes mean nothing.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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
Mobile Verification
The trap this exists to escape
Most browser-automation tools drive a tab that is never visible and never focused:
document.visibilityState -> "hidden"
document.hasFocus() -> falserequestAnimationFrame is throttled to ZERO in a hidden tab. Anything animated silently never moves. And scroll-behavior: smooth — which most modern sites set on html — animates via rAF.
So on a page that works perfectly for a human, that harness reports:
location.hash = '#target' -> no movement
element.scrollIntoView() -> no movement
window.scrollTo({behavior:'smooth'}) -> no movement
a REAL TRUSTED MOUSE CLICK on an anchor -> no movementI concluded "the whole site cannot be scrolled," closed a good PR, wrote an alarming brief, burned three builds, and told the owner production was broken. A five-second check on his phone proved it worked fine.
Know what your harness is structurally blind to. Mine could not observe motion — not because of a bug, but because of what it was.
The one-line discriminator
Before you ever conclude "the page can't scroll":
window.scrollTo({ top: 1000, behavior: 'instant' }); // bypasses rAF entirelyIf INSTANT works and SMOOTH doesn't, your tab is hidden and your test is invalid. Not the page.
The fix: Playwright
Playwright's pages are genuinely visible and focused. Verified:
visibilityState: "visible" hasFocus: true
rAF ticks/300ms: 21 ← animations actually run
viewport: 393x659 touch: true dpr: 3 ← real iPhone 15 Pro
REAL TAP → scrollY 0 → 2695, target section landed at top: 0 ✅npm i playwright && npx playwright install chromium
node scripts/mobile-check.mjs "<url>" --tap ".cta-bar" --expect-visible "#target"scripts/mobile-check.mjs checks in one pass:
- Horizontal overflow — names the offending elements (ignores intentional
overflow-x:autoscrollers) - Touch targets under 44px — Apple's floor; we shipped 40px more than once
- The tap outcome — did the page scroll, did the expected element land on screen
- Console errors
- A screenshot
And it self-aborts if its own tab isn't visible or rAF isn't ticking. A harness that cannot see motion should say so, not report a false failure.
Prove your harness can FAIL
The most important step, and the one everybody skips:
# assert something you KNOW is false
node scripts/mobile-check.mjs "<url>" --tap ".cta-bar" --expect-visible "#something-far-away"
# must print 🔴 FAILA harness that has never failed cannot be trusted. If it can't fail on a false assertion, its passes mean nothing. (My first hook test "passed" because the command never ran at all.)
When to use which tool
| Job | Tool |
| Read the DOM, grep the HTML, inspect computed styles | Any automation / extension. Fine. |
| Anything a USER DOES — tap, scroll, animate, transition | Playwright. |
| "Does this feel right" / "does this look cramped" | Ask a human. Five seconds. |
That last row is not a defeat. Some things a harness is structurally blind to — visual taste, feel, motion on a real device. An hour of an agent chasing a ghost is worth strictly less than one five-second look at a phone.
Higher-fidelity options, if you need them
- iOS Simulator (free, ships with Xcode) — real WebKit/Safari, not Chromium. Use when the bug smells iOS-specific:
-webkit-quirks, Safari's viewport units, momentum scrolling, input zoom. - BrowserStack / LambdaTest / Sauce Labs (paid) — real physical devices. Use when a bug only reproduces on hardware.
Playwright's Chromium emulation covers the overwhelming majority of layout and interaction bugs, and it's the only one of the three that's free and fast enough to run on every single build.
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.

