agentleFS
duanebesterduanebester/gooey/.agents/skills/previewing-gooey/SKILL.md

previewing-gooey

Skill692 starsChanged 4 days agoMITView on GitHub

Builds and runs Gooey examples in a headless Wayland session and captures reviewable screenshots. Use when asked to run, preview, inspect, or screenshot a Gooey app in an Amp orb, or to test Gooey on GNOME, Wayland, or Linux (input, IME, resize, every example on both Wayland backends).

  • In this repositoryOne other file it gives its agents.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 1 is distinctive, and 2 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.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Every example shows a fully drawn first frame.

    line 51
  2. Very specific:

    Expected on GNOME: no title bar, because mutter offers no server-side decorations and Gooey draws none.

    line 75
  3. Very specific:

    No screenshot, or input does nothing: check amp orb service logs gnome-shell and amp orb service status gnome-pipewire.

    line 91

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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

Previewing Gooey

Two workflows, both headless and rendered with Mesa's software Vulkan (llvmpipe):

  • Quick preview: the todo example under Weston, one screenshot.
  • GNOME check: a GNOME Shell (mutter) Wayland session with real input, IBus, and screenshots, for testing behaviour on Linux.

Quick preview (Weston)

.agents/skills/previewing-gooey/scripts/capture-todo

The script builds zig build todo with the checked-in SPIR-V shaders, starts Weston and Gooey as supervised orb services, and captures .amp/in/artifacts/gooey-todo.png. Pass a repository-relative output path to choose another name. If the screenshot is blank, check amp orb service status gooey-app and amp orb service logs gooey-app. Stop it with amp orb service stop gooey-app and amp orb service stop gooey-wayland.

GNOME check (mutter on Wayland)

.agents/skills/previewing-gooey/scripts/gnome-session start
.agents/skills/previewing-gooey/scripts/gnome-check        # vulkan | cpu | all (default)
.agents/skills/previewing-gooey/scripts/gnome-session stop

gnome-session start runs GNOME Shell headless on a 1280x800 virtual monitor (--wayland --no-x11), with a session bus, PipeWire, and WirePlumber, all as orb services. It adds IBus Anthy as a second input source and writes the client environment to /tmp/gnome-rt/env.

gnome-check builds every example for wayland_vulkan and wayland_cpu. For each one it launches it, captures a frame, closes it with Alt+F4, and requires exit status 0 and no panic, assertion, or capacity error on stderr. It then maximizes todo and checks typing, Backspace, Home and End, Shift, two Adds (it requires both rows), and Japanese input: preedit, conversion, the candidate window, and commit. It prints ok or FAIL per check and exits nonzero on any failure.

Screenshots go to .amp/in/artifacts/gnome/: one per example, a vulkan-all.png and cpu-all.png contact sheet, and todo-edit.png and todo-ime-*.png. A frame that draws wrong still passes, so inspect them with view_media:

  • Every example shows a fully drawn first frame.
  • todo-edit.png lists "XBuy milk!" and "Walk dog".
  • todo-ime-preedit.png shows underlined kana, and todo-ime-candidates.png shows the candidate list directly under the field.

Driving a session by hand

source /tmp/gnome-rt/env
./zig-out/bin/gooey-counter > /tmp/counter.log 2>&1 &
.agents/skills/previewing-gooey/scripts/gnome-input \
    "key Super_L+Up; sleep 1; click 640 400; type Hello!; shot /tmp/screen.png"

gnome-input takes ;-separated commands: move X Y, click [X Y], drag X0 Y0 X1 Y1, scroll STEPS, key A+B, down KEY, up KEY, type TEXT, shot PATH, and sleep SECONDS. Coordinates are screen pixels; GNOME's top bar is 32 px. The headless seat has no input devices of its own, so each call first adds its virtual pointer and keyboard with a zero motion and a Shift tap; without that, its first click or key reaches no window. Run a short-lived app as a background child, as above, so its exit status stays readable. Run long-lived servers as orb services.

What the session cannot show

  • Expected on GNOME: no title bar, because mutter offers no server-side decorations and Gooey draws none. Alt+F4 closes the window.
  • Not exercised on GNOME 43: the wp_cursor_shape_v1 cursor path, which mutter added in GNOME 48. Hover cursors here come from Gooey's XCursor theme fallback.
  • Not covered: hardware GPUs, fractional scaling, and physical displays. Test those on a real Linux machine.
  • Fonts differ: text layout uses the orb's fonts (DejaVu Sans), which are wider than macOS and Windows fonts.

Troubleshooting

  • Typed text arrives as plain Latin: IBus is still on the US layout. The input-source switch (Super+Space) works only while a Gooey text field has focus; ibus engine prints the active engine.
  • ibus list-engine lacks anthy: rerun gnome-session start; it rebuilds IBus's engine cache.
  • No screenshot, or input does nothing: check amp orb service logs gnome-shell and amp orb service status gnome-pipewire. Absolute pointer motion needs PipeWire.

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.