MGallo-Code

ui

Michael's UI workflow and rules. Use for any change someone can see (building, changing, mocking or reviewing a UI, page, component, review page, chart or rendered document), from the start of the work through to calling it done.

MGallo-Code 0 Updated 2d ago

Resources

9
GitHub

Install

npx skillscat add mgallo-code/dotfiles/ui

Install via the SkillsCat registry.

SKILL.md

UI workflow

Run these steps in order for any visible change. Skip it only for backend, CLI or library work with no visible output.

1. Read the rules that apply

  • Always read rules.md: the 25 core rules, wording and labels first. They are the corrections Michael makes most.
  • Then read each topic file whose trigger matches the work:
    • core.md (APP-): any app with more than one screen.
    • components.md (COMP-): forms, lists, tables, dialogs, menus, pickers, editors, media.
    • touch.md (TOUCH-): anything on a phone or tablet, or taking touch input.
    • desktop.md (DESK-): anything used on a desktop or laptop screen.
    • assistant.md (CHAT-): a chat or AI-assistant interface.
    • personal-tools.md (SOLO-): a tool only Michael uses, or a mock or review page for him. Never a client product or KeepTheCall.
    • process.md (PROC-): how to plan, mock, review and report UI work with him.
  • Taste, from his own words:
    • clean simplicity with selective premium touches, not minimalism for its own sake;
    • start simple and add flair selectively;
    • color-match to the images or photos in use;
    • think about the end user (older users need clear headings);
    • keep spacing, borders and widths consistent;
    • real call-to-action buttons over subtle links.
  • Rule format: FORMAT.md. A rule's Touches line names a rule it interacts with, and which wins.

2. Mock when there is a real choice

When a visible change has several reasonable designs, show the options at 390 and 1280 px before building (PROC-01). Use real-looking sample data, never stubs (UI-10). Clear fixes skip the mock: build them, then show before and after.

3. Build

Use the app's shared components and style variables (UI-13). Build each change for every device the product runs on (UI-24).

4. Verify live

  • Drive the running app or page with Playwright (the connected Playwright MCP, not chrome-devtools): load it, exercise the change, take screenshots at phone and desktop widths, and read the console and network for errors.
  • Confirm that it renders and works, not just that the code looks right.

5. Review against the rules

Walk the change, on screen, against rules.md and the topic files you read: every visible string first, against UI-01 to UI-08 and the no-redundant-copy rule; then the rest. Fix every failure you can.

6. Show him, then report

  • Open the result on his screen where possible: start the dev server and open <localhost url>. Screenshots or a visible Playwright session will do when that isn't possible.
  • Do this before any PR or "done".
  • Close with the three-part report: done (and where to try it), not done, needs his decision. Anything from step 5 left unfixed goes under "not done".