key-indicator-helper

---

name: key-indicator-helper description: Render a "press this key" indicator on screen — a glyph (letter / SPACE / SHIFT / LMB) painted on a key-shaped tile (a keycap, square, or pill texture). Use when a project needs keypress hints, input prompts, a controls overlay, a tutorial "press W/A/S/D" row, or any "show the player which key to press" UI. Owns the hard part — placing a glyph on a beveled cap's TOP FACE (not its box center), sizing the font and cap so it reads as a label, swapping square↔wide by label length, and an optional looped press-down call-to-action. Genericized to any project: it references no specific keycap art — you supply the texture and measure it with the /see skill. Ships a drop-in C++ UMG widget template and the layout math a consumer uses to size it. user-invocable: true allowed-tools:

• Bash
• Read
• Glob
• Grep
• Skill
• Write
• Edit
• AskUserQuestion

---

key-indicator-helper — put a glyph on a key, correctly

You need to tell the player "press F" (or W, SPACE, SHIFT, LMB). The right way to do that is a rendered key indicator — a small key-shaped texture (a keycap / square tile / pill) with the key's glyph painted on its top face — not bracketed plain text like [F]. This skill owns the part that is deceptively hard to get right, gives you a drop-in widget that encapsulates it, and tells you exactly how to size and verify it. It references no project-specific art — you bring the texture; the method is universal.

Why this is harder than it looks (the intellectual core)

A key-shaped texture is a small picture with a 3D-ish bevel/skirt and a flat top face. Four traps sink the naive "drop a TextBlock in the middle" approach:

1. The drawable face is NOT the box center. The bevel/skirt pushes the flat top face up (and often slightly sideways). Center the glyph on the box and it sits on the skirt or below the face. You must anchor to the measured face center — a (cx, cy) fraction you get from the /see skill, confirmed by eye on its annotated PNG.
2. A capital letter's ink sits high in its line box. Even anchored to the face center, the glyph reads as floating. Nudge the anchor down ~3% of the box so the ink (not the line box) centers on the face.
3. The glyph is a label, not a fill. Size the font to ~⅓ of the cap-box height (FontOfBox ≈ 0.30). Bigger and it looks like a filled tile; the cap should read larger than its glyph.
4. Transparent padding lies about size. The texture canvas usually has transparent margin around the opaque cap. To make the visible cap match a target on-screen height, divide by the opaque-body fraction (trimmed_height ÷ canvas_height, from /see) when you compute the box height — otherwise the cap renders smaller than you asked.

Plus two structural facts: single-char vs multi-char keys want different art (a square cap for F, a wide cap for SPACE/SHIFT, with its own face center), and the cap should be caller-sized — the consumer computes a box height from its own layout band and the cap scales everything (font, press depth) from it.

The shipped template (templates/KeyIndicatorWidget.{h,cpp}) encapsulates all of this. Your job is to measure your art, supply it, and size the widget — the traps are already handled.

The workflow

1. Get / pick the key art   → a square cap texture (+ optional wide cap for SPACE/SHIFT)
2. /see it                  → face-center (cx,cy) fraction + opaque-body fraction, per texture
3. Import it                → asset_textures_import into /Game/...
4. Drop in the template     → copy KeyIndicatorWidget.{h,cpp} into the project's C++ module
5. Wire one cap             → SetCapTextures + SetFaceCenters + SetGlyph, add to a panel
6. Size it                  → SetCapSize(boxHeight) each layout pass (math below)
7. Verify                   → PIE + editor_screenshot; the glyph sits ON the face, crisp

1–2. Measure the art with /see (do not eyeball)

Run the /see skill on each cap texture:

python .claude/skills/see/scripts/see.py path/to/keycap_square.png
python .claude/skills/see/scripts/see.py path/to/keycap_wide.png

From its JSON + annotated PNG, record per texture:

• Face center (cx, cy) — start from centroid / interior_solid, then confirm the top-face center by eye on the *.see.png (the bevel offsets it from the box center — this is exactly the see "face-center caveat"). This is the most important number.
• Opaque-body fraction — bbox_solid.trimmed_size[1] ÷ image.height (the visible cap's height as a fraction of the canvas). Used in the sizing math.
• Wide aspect — image.aspect of the wide cap (the template reads this from the texture automatically, but you'll want it for layout).

A square cap with a beveled top face typically lands near (0.50, 0.41); a wide cap near (0.50, 0.46). Treat those as a sanity range, not a substitute for measuring yours.

No key art yet? Any front-on keycap/square-tile PNG with a transparent background works (a light cap with a dark face for dark glyphs, or a dark cap for light glyphs). To generate an on-theme cap, use the /icon skill (--theme to match your game) — it needs an OpenAI key (the /onboard skill sets it up, or you're prompted on first generation). Or hand-author / source the PNG. Either way, /see + this widget take over once you have it.

3. Import the texture

asset_textures_import (MCP asset domain) brings the PNG into /Game/.... Keep the import path stable — the widget loads it by path or you pass the UTexture2D* in.

4–6. Drop in the widget, wire it, size it

Copy templates/KeyIndicatorWidget.h and .cpp into the consuming project's C++ module (Source/<Module>/UI/), then make two edits:

• Replace KEYINDICATOR_API with the project's module API macro (e.g. HOVERBALL_API).
• Nothing else is required — textures, face centers, and color are passed at runtime.

Wire one cap from your HUD / widget:

UKeyIndicatorWidget* Cap = CreateWidget<UKeyIndicatorWidget>(this, UKeyIndicatorWidget::StaticClass());
Cap->SetCapTextures(SquareTex, WideTex);                 // WideTex may be null
Cap->SetFaceCenters(FVector2D(0.50f, 0.41f),             // ← YOUR measured square face center
                    FVector2D(0.50f, 0.46f));            // ← YOUR measured wide face center
Cap->SetGlyphFromKey(EKeys::F);                          // or SetGlyph(FText::FromString("F"))
MyPanel->AddChildToHorizontalBox(Cap);                   // any panel slot

Then drive its size every layout pass from your own layout. The cap is a USizeBox; you tell it a box height and it scales the font and press depth from that.

The sizing math (consumer side)

Compute the cap-box height from the space you have, then call SetCapSize. Re-run on viewport resize (in NativeTick, when MyGeometry.GetLocalSize().Y changes — that's the DPI-corrected viewport height for a viewport-added widget; reliable, unlike GetViewportSize at construct):

void UMyPromptWidget::NativeTick(const FGeometry& G, float Dt)
{
    Super::NativeTick(G, Dt);
    const float H = G.GetLocalSize().Y;
    if (H > 1.f && !FMath::IsNearlyEqual(H, LastLayoutH, 0.5f)) { ApplyLayout(H); LastLayoutH = H; }
}

void UMyPromptWidget::ApplyLayout(float ViewportH)
{
    // Want the VISIBLE cap to be ~6% of viewport height. Divide by the opaque-body
    // fraction (from /see) so the transparent padding doesn't shrink the visible cap.
    constexpr float VisibleCapOfViewport = 0.06f;   // tune to taste
    constexpr float OpaqueBodyFrac       = 0.695f;  // ← from /see (trimmed_h ÷ canvas_h)
    const float CapBoxH = (ViewportH * VisibleCapOfViewport) / OpaqueBodyFrac;
    Cap->SetCapSize(CapBoxH);
}

If the cap sits inside a band/plaque (a prompt strip) rather than the raw viewport, scale relative to the band height instead: CapBoxH = (BandPx * KeyFillOfBand) / OpaqueBodyFrac, where KeyFillOfBand (~0.4) is how much of the band's height the visible cap should fill, and BandPx comes from your plaque layout. Same idea — always divide by the opaque-body fraction.

Square vs wide caps

SetGlyph switches automatically: a label longer than one character (SPACE, SHIFT, CTRL, ENT, LMB) uses the wide texture if you supplied one (with the wide face center and the texture's own aspect); single characters use the square cap. If you pass no wide texture, multi-char labels stay on the square cap (they'll just be cramped — supply a wide cap for a clean result). KeyLabel(FKey) produces the terse cap text (SPACE, SHIFT, LMB, ENT, or a single uppercase letter) and is static so a controls table can share it.

The press-down call-to-action (the "tutorial" pattern)

For a tutorial / onboarding prompt that says "now press these," enable the looped press animation and phase-offset a row of caps so they ripple instead of pressing in unison:

// A row of caps: W A S D, rippling left-to-right.
const float RippleStep = 0.18f;   // seconds of phase offset per cap
for (int32 i = 0; i < Keys.Num(); ++i)
{
    UKeyIndicatorWidget* Cap = CreateWidget<UKeyIndicatorWidget>(this, UKeyIndicatorWidget::StaticClass());
    Cap->SetCapTextures(SquareTex, WideTex);
    Cap->SetFaceCenters(SquareFace, WideFace);
    Cap->SetGlyphFromKey(Keys[i]);
    Cap->SetPressCTAEnabled(true);
    Cap->SetPressPhaseOffset(i * RippleStep);     // ripple, not unison
    Row->AddChildToHorizontalBox(Cap);
}

The animation is a code-driven render-transform (down → hold → release → pause, looped) — no UWidgetAnimation asset. A static prompt (e.g. an interaction "[F] Open") simply leaves the CTA off (the default).

Showing the player's bound (rebindable) key

SetGlyphFromKey(FKey) shows whatever key you hand it. To show the live bound key (so a rebind reflects in the prompt), resolve the FKey from your project's input/settings system first, then pass it in — e.g. read the Enhanced Input mapping for the action, or your settings subsystem's override, falling back to the default. The widget deliberately does not know your binding system; keep that resolution in the project and feed it the resulting FKey.

Alternative route: pure MCP widget tools (no C++ / no build)

The C++ template is preferred — it paints crisp Slate fonts the immediate-mode HUD canvas cannot, and encapsulates the math. But if a project can't take a C++ widget, the same method works through the MCP widget domain (widget_create, widget_add_child, widget_set_property): build a Canvas/Overlay with an Image (the cap texture) and a TextBlock (the glyph) as siblings, and set the TextBlock's anchor to the measured face center + alignment (0.5, 0.5) + the down-nudge, and its font size to ~⅓ of the cap box. The hard numbers (face center, opaque-body fraction) come from /see either way — only the wiring differs. Lose the press animation unless you script it via a widget animation.

Verify (do not skip)

1. Build, run PIE, editor_screenshot (GUI mode for render — see tests/run.ps1 --ue-mode=gui).
2. Look at it like a critic. Is the glyph centered on the top face, not the skirt? Is it crisp (real Slate font, not blurry HUD-canvas text)? Does the visible cap match the size you intended (if it's small, your opaque-body-fraction divide is missing or wrong)? For a wide cap, is the multi-char label centered on the bar? If anything's off, re-measure the face center with /see and re-look — almost every defect traces to a wrong face-center fraction or a missing opaque-body divide.

Boundaries & honesty

• No project-specific art, paths, or lore. This skill never names a particular keycap asset. The technique and the template are universal; the texture + its measured fractions are the project's to supply.
• Measure, don't guess. Every face-center and size number comes from /see (run + verified by eye), not from a remembered constant. The defaults in the template are a starting estimate for a typical beveled cap and are explicitly there to be overwritten.
• /see is a hard dependency for correctness — it's the measurement half of this skill. This skill is the wiring half.