Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →hashgraph-online/ui-theme-qa
Use this skill to find and fix undisciplined Control-node UI in a Godot project — spacing that follows no ladder, font sizes with no modular scale, near-duplicate colours, and text that fails WCAG contrast — by measuring the live editor with the `hera` CLI and snapping theme tokens to a reference corpus. It also reports (without changing) inert container wrappers and decorative nodes. Triggers: "clean up the UI spacing/type/contrast", "the UI values look arbitrary", "tidy the Godot UI", "check U
Passed 9 of 15 test cases with the skill, and 2 of 15 without it.
and 13 more in the full report
10 fixed3 passed either way9 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 10 test cases the skill fixed. Same model, same prompt, with and without the skill.
What are the exact terminal commands used to launch and terminate the game to capture the confirmation gate screenshot for the UI defect pipeline?
| Checked | Without skill | With skill |
|---|---|---|
| Executes hera run --current --wait to launch the game | Missed | Passed |
| Executes hera stop --wait to terminate the game after analysis | Missed | Passed |
The instructions your model reads when the skill loads.
Mechanically remove statistical design defects from a live Godot editor, over
Control nodes. It does not redesign: it snaps theme tokens to a reference
corpus and never touches copy, information order, or layout structure.
Six areas, split by what they may do:
| areas | effect | |
|---|---|---|
| enforcing | spacing · type-scale · color · contrast | set theme_override_* (undoable) |
| report-only | containers · decoration | proposals in the report; no mutation |
The report-only pair would have to delete or re-parent nodes, which cannot be
decided mechanically — a node that looks decorative may be a divider, and
flattening a wrapper breaks the $Path/To/Node references scripts rely on. They
measure and surface; a human decides.
Why a pipeline, not one pass: one context holding every rule while fixing a large UI lets loud items crowd out quiet rules, which get silently dropped. So inspection and enforcement are split per area, and findings live in files. The orchestrator (this context) stays thin — it holds routing only; the detailed rules and findings live in the reference doc and the findings files.
Rules SSOT: references/ui-theme-areas.md. Values: references/reference-corpus.md. Read only the area section you are working; do not load all rules into one context.
hera status — confirm one live editor (the live-editor skill covers
setup). If multiple, pass hera --instance <pid> to every mutation.hera guidance ui — respect the returned UI mode.hera --ids scene tree (edited scene) — this
is the static surface the checks read.test cases passed, vs 2 / 15 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.