agentleFS
MicrosoftKnown publishermicrosoft/PilotSwarm/.github/skills/portal-ux-testing/SKILL.md

portal-ux-testing

Skill17 starsChanged 9 months agoMITView on GitHub

Test a browser portal UX change in every layout mode it can meet - reading column on and off, live and after a reload, desktop and phone, Chromium and WebKit, with and without a canvas. Load before you call a portal layout, chat, session-list or canvas change done.

  • Reads credentials

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    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
  1. Distinctive:

    Not every change reaches every axis.

    line 27

≈ “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.

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

Portal UX testing

A portal change that looks right in one mode often breaks in another. Most layout regressions found by the owner were in a mode nobody tested: the reading column turned off, a canvas that popped up during a live turn, a phone, or Safari. Test the change in each mode below that it can reach, and say which ones you covered.

The mode matrix

AxisTest bothWhy
Reading column (Settings > Appearance)chatReading.on true and falseWith it on, chat content is held to a fixed column. With it off, wide content (tool cards, code, tables, artifact cards) is free to size against the pane, and grid tracks see different minimum widths. A fix proven with one setting proves nothing about the other.
Live vs reloadedDuring a live turn and after a page reloadLive state (a canvas that pops up mid-turn, streaming cards, message previews) takes different code paths from state hydrated from the profile and history. "Broken until refresh" means only the reloaded path was tested.
Settings round tripBefore and after opening and closing Settings (the Admin Console)The console replaces the whole workspace and puts it back. Anything bound to the old elements (measurements, observers, refs) goes stale until a reload. Changing a setting such as the reading column always takes this path, so test a setting by changing it live, not only by seeding it.
Session switchingSwitch to another session and back, more than oncePer-session view state (canvas open, remembered canvas width, scroll anchor) is saved on leave and restored on return.
CanvasClosed, open at default width, and a remembered width wider than the window can giveRemembered widths are per session and can exceed the window. Nothing else may pay for them: the session list keeps its width and chat keeps its floor.
Device classDesktop (1280, 1440, 1920) and phone (390x844)The phone has its own workspace: the session list is a slide-over, Settings > Appearance carries phone-only rows. Desktop settings such as the reading column must not leak to the phone.
BrowserChromium and WebKitiPhone and Mac Safari are WebKit. A spec opts in with test.use({ browserName: process.env.PS_TEST_BROWSER || "chromium" }); run it again with PS_TEST_BROWSER=webkit. Phone device presets (devices["iPhone 14 Pro"]) already run WebKit.
ThemeOne modern theme and one retro theme (for example cobalt2 and win95)Retro themes change borders, fonts and control sizes, and so the geometry.

Not every change reaches every axis. A phone-only change does not need the reading column; a session-list change does. When you skip an axis, say why.

Setting the modes in a spec

Seed profile settings through the profile route, as chat-rail-fit.spec.mjs does, and stub the settings write so a test never stores a profile:

await page.route("**/api/v1/**", route => {
    const url = new URL(route.request().url());
    if (url.pathname.endsWith("/me/profile")) return route.fulfill({ json: { ok: true, result: { isAdmin: false, profileSettings: {
        themeId: "cobalt2",
        chatReading: { on: reading, width: 1100, align: "left", pad: 20 },
    } } } });
    if (url.pathname.endsWith("/me/profile/settings")) return route.fulfill({ json: { ok: true, result: {} } });
    return route.fallback();
});

Loop the test over both values rather than writing two specs:

for (const reading of [false, true]) test(`... (reading column ${reading ? "on" : "off"})`, async ({ page }) => { /* ... */ });

For live behaviour, deliver the event the way production does (the events route or the live plane) after the page has loaded, measure, then reload and measure again. Assert the two agree.

What to assert

Assert properties, not pixels: computed styles and getBoundingClientRect() geometry. For layout stability, record the measurement in each state and assert it holds, for example the session list width across canvas and non-canvas sessions, live and after reload, with the reading column on and off. A fix needs a test that fails without it: run the new test against the old code once and report that it failed.

On a deployed test environment

Specs run against the stub server. Before you call a layout change ready, also check the deployed build in a real browser with the owner's own profile shape (per-session canvas widths, reading column setting, long session list), in both reading-column modes, live and after reload. Route the profile-settings write to a stub unless the run is meant to change the profile.

Report

State the modes covered in the result, for example:

ModeCovered
Reading column on / offboth
Live / reloadedboth
Desktop / phonedesktop only (change is desktop-only)
Chromium / WebKitboth

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.