@funpixel/tools 0.1.0 → 0.1.1

package/plugins/claude/skills/3d/references/levels.md

@@ -1,6 +1,6 @@
 # Quality levels: one source, a cut per device
 
-Held Note decimated every model once for every device (the station kit 150,000 → 4,000
+One game decimated every model once for every device (the station kit 150,000 → 4,000
 triangles, characters → 13,000, props → 1,200–3,000, textures 4096² → 512–1024²). Next to its
 target frames it read flat and melted on a desktop that had headroom to spare, and the untouched
 generations survived only in a git-ignored temporary folder. Levels are the other way to do it:
@@ -40,7 +40,7 @@ failure: the level is the most the source allows at that quality.
 Where the source lives is your call. A generation made with `gen model3d` keeps the provider's
 untouched GLB in `.funtools/sources/models/` (never published); the platform keeps its own copy
 beside the shipped one. Cut from that, never from a file you already decimated: a level cut from
-Held Note's shipped walker (18K triangles) reached only 16K on phone, because the detail it needed
+its shipped walker (18K triangles) reached only 16K on phone, because the detail it needed
 to trade was already gone.
 
 A `hero` generation (`style: "hero"`) is the source this is built for, cut down per device
@@ -67,7 +67,7 @@ saver or a low GPU gets `phone`; `high` needs a GPU known to be high
 (`levelSignals(renderer.getContext())` reads it) and 8 GB; everything else `desktop`. A phone never
 downloads the desktop set. No manifest (a model never cut) loads the plain file, as before.
 
-## Measured on Held Note's sources (29 Sep 2026)
+## Measured on that game's sources (29 Sep 2026)
 
 Release build of the engine on a shared M-series laptop; the time is mostly KTX2 texture encoding.
 

package/plugins/claude/skills/3d/references/light-and-places.md

@@ -1,7 +1,7 @@
 # Light, dressing and places: directing the world
 
-A world reads as shapes when it is evenly lit, sparsely dressed and the same everywhere. Gallery
-Heist's plan was dark and moonlit, with spot-lit art and deep shadow, and its first build came back
+A world reads as shapes when it is evenly lit, sparsely dressed and the same everywhere. One heist
+game's plan was dark and moonlit, with spot-lit art and deep shadow, and its first build came back
 bright and evenly lit. Silt & Signal, by contrast, puts 42k instanced flowers along a clear road and
 gives each of its seven places its own light. The 3D starter (`funtools init --framework three-world`)
 directs all three from data, so you change values instead of writing a renderer:
@@ -116,7 +116,7 @@ palette (`sky-top`, `sky-horizon`, `ground`, `ground-2`, `stone`, `accent`, `glo
 (`grass`, `sand`, `cobble`, `flagstone`, `snow` or `moss`); its kit; its landmark; its road; its
 `density` target; and its `target` picture when the plan has one.
 
-Gallery Heist's spec is the model for a row: each room has a size and shape, a material, an era, a
+One heist game's spec is the model for a row: each room has a size and shape, a material, an era, a
 light and a colour, and differs from the others in all five. For example:
 
 - "Long Gallery, oxblood velvet, cool moonlight from the lunette against warm gold picture lights"

package/plugins/claude/skills/3d/references/painted-target.md

@@ -1,19 +1,19 @@
 # From a painted target to a 3D scene that holds up
 
-Why this exists: Blue Freight was built twice from one plan (1 Oct). In both builds the title and
+Why this exists: one game was built twice from one plan (1 Oct). In both builds the title and
 menus reached their painted targets, and the 3D walk stayed a flat, empty plane under a plain
 sky: a tech demo. In one build the cut-out hero's legs and arms drifted off the body as she
 walked. The walk is the first minutes of play, the stretch that decides whether people stay (the
-hook). Held Note shows the way that worked: every look pass ended in fidelity strips at fixed
-framings, today beside the target, and the next pass worked on the largest gap. Saltwake (boats of
-boxes and cones handed back as finished) and Fold Room (57 code primitives) show what happens
+hook). Another game shows the way that worked: every look pass ended in fidelity strips at fixed
+framings, today beside the target, and the next pass worked on the largest gap. A racer (boats of
+boxes and cones handed back as finished) and a puzzle game (57 code primitives) show what happens
 when nobody puts the game beside the picture.
 
 ## Read the target first (five minutes, written down)
 
 Before any code, measure the target frame and write the answers beside it in the GDD's place row:
 
-| Read | Example (Blue Freight, f3_walk) |
+| Read | Example (one game, f3_walk) |
 |---|---|
 | Horizon height, share of frame from the top | 0.42 |
 | Hero height, share of frame | 0.22, standing left of centre |
@@ -93,7 +93,7 @@ point at a closed gap in its strip did not close it.
 
 # A walk that reads as a person, in one piece
 
-The owner, after Blue Freight: "better walking animations, not like a robot; smooth movement like
+The owner, after that game: "better walking animations, not like a robot; smooth movement like
 us humans." A robot walk is limbs swinging at a constant speed about fixed points while the body
 glides. A person's walk is the body falling forward and catching itself.
 
@@ -150,7 +150,7 @@ size on screen wins, whichever took less work.
 ## Many painted cut-outs, one draw call
 
 A painted world of props, creatures and folk drawn as one `THREE.Sprite` each costs a draw call
-each: the Blue Freight build reached 164 against its budget of 150 with about 200 cut-outs. The 3D
+each: one build reached 164 against its budget of 150 with about 200 cut-outs. The 3D
 world starter's `src/cutouts.ts` draws every cut-out that shares an atlas as one instanced mesh, so
 300 cut-outs are one draw call. Cut the figures with `npx @funpixel/tools cut sheet.jpg` (clean
 alpha, no paper or floor patches), build the atlas in the page with `atlasFrom([...pngs])` (or use

package/plugins/claude/skills/3d/references/three-loading.md

@@ -48,8 +48,8 @@ root.traverse((o) => { const m = (o as THREE.Mesh).material as THREE.MeshStandar
 
 A free HDRI, when the game has one, replaces `RoomEnvironment` as both lighting and sky.
 
-Judge the result in the game's own light, not the generator's preview. Held Note (a dark three.js
-horror game, 29 Sep 2026) tuned every material in its loader rather than in the files: the hero's
+Judge the result in the game's own light, not the generator's preview. A dark three.js
+horror game (29 Sep 2026) tuned every material in its loader rather than in the files: the hero's
 suit to metalness 0.3 (it read as chrome under the torch), the crew's cloth to metalness 0 and 72%
 albedo (it burnt white in the beam). Its generated textures read near-black, and lifting the
 enemy's albedo ×1.9 made it go pure white a few metres into a spotlight: a colour above 1 has
@@ -169,13 +169,13 @@ function floorOffset(mixer: THREE.AnimationMixer, clip: THREE.AnimationClip) {
 
 Optimized GLBs are often quantized: positions stored as normalized int16, the real scale carried in
 the node. Baking a node transform into that geometry (`geometry.applyMatrix4`) clamps every vertex
-to ±1; Held Note's 3 m wall modules stood 1 m tall for a whole pass. Keep transforms in the object
+to ±1; one game's 3 m wall modules stood 1 m tall for a whole pass. Keep transforms in the object
 or instance matrix. The same quantization moves each mesh's scale into its node, so a part scaled
 at runtime (a gauge's fill) has that scale baked back first, or `scale.y = fraction` is not literal.
 
 A skinned mesh keeps `frustumCulled = false` (its bind-pose bounds lie), so nothing culls a crowd
 of them: every one animates and casts shadows everywhere. Cull them by distance and screen yourself
-(skill `fp:art`, its frames reference, has Held Note's numbers).
+(skill `fp:art`, its frames reference, has one game's numbers).
 
 Generated models usually face +Z, but not always. Wrap each in a group; yaw the inner model once
 at load (a vehicle whose posed size has x greater than z was authored sideways), and keep steering

package/plugins/claude/skills/3d/references/unity.md

@@ -1,17 +1,17 @@
 # 3D models from generator to Unity
 
-A 3D model can fail in Unity without an error. glTFast (the glTF importer, package `com.unity.cloud.gltfast`) decodes textures embedded in a GLB to raw RGBA32, 4–16 MB each, and imports a web GLB packed with meshopt or WebP as nothing. That made the 3D games 3–5× bigger than the 2D ones: Noodle Nights was 386 MB before this pipeline and 106 MB after. This skill takes a model from the generator to a compressed, collidable asset.
+A 3D model can fail in Unity without an error. glTFast (the glTF importer, package `com.unity.cloud.gltfast`) decodes textures embedded in a GLB to raw RGBA32, 4–16 MB each, and imports a web GLB packed with meshopt or WebP as nothing. That made the 3D games 3–5× bigger than the 2D ones: one idle game was 386 MB before this pipeline and 106 MB after. This skill takes a model from the generator to a compressed, collidable asset.
 
 ## Tools
 
-The studio's `com.joydle.core` package carries the texture import rule, `TextureBudget` (fp:unity-game, `references/studio-packages.md`): a texture named `model__basecolor.png` (or `__normal`, `__mr`, `__emissive`) under an `Art/` or `Models/` folder imports as ASTC 6×6 with mipmaps, sRGB only for base colour and emissive, at most 512 under `Food/`, `Props/` or `Characters/` and 1024 elsewhere. Not shipped by the kit or the package: the Unity starter, its checks and the studio scripts (`blender_prep.py`, `render_glb.py`, a collision audit, a size audit). `TextureBudget` imports `__normal` maps as Default (field report, 10 Oct). Do the rest with what exists:
+Not shipped by the kit: the Unity starter, its checks and the studio scripts (`blender_prep.py`, `render_glb.py`, a collision audit, a size audit). Do the rest with what exists:
 
-- Generate into the project: `npx @funpixel/tools gen model3d "<subject>" --image-asset concept.png --unity .` (after `npx @funpixel/tools login`) asks for the engine's own GLB (`--raw`: glTFast reads no Draco, meshopt or KTX2 without three more packages), splits it into `.gltf` + `.bin` + `model__basecolor`/`__normal`/`__mr` textures in `Assets/_Game/Generated/Models/<name>/` (the studio's `glb_split.py`, ported), and writes an import rule that sets those textures once, ASTC 6×6 at most 1024, sRGB only for colour, and keeps `__normal` a NormalMap over `TextureBudget`. Checked on 6000.6 with a Meshy and a Hunyuan model: meshes and textures import. The studio's `gen_3d.py` calls fal with a `.env` key: no metering, no cost log.
+- Generate into the project: `npx @funpixel/tools gen model3d "<subject>" --image-asset concept.png --unity .` (after `npx @funpixel/tools login`) asks for the engine's own GLB (`--raw`: glTFast reads no Draco, meshopt or KTX2 without three more packages), splits it into `.gltf` + `.bin` + `model__basecolor`/`__normal`/`__mr` textures in `Assets/_Game/Generated/Models/<name>/` (the studio's `glb_split.py`, ported), and writes an import rule that sets those textures once, ASTC 6×6 at most 1024, sRGB only for colour, and keeps `__normal` a NormalMap. Checked on 6000.6 with a Meshy and a Hunyuan model: meshes and textures import. The studio's `gen_3d.py` calls fal with a `.env` key: no metering, no cost log.
 
 - Prepare in Blender through the kit's bench: `npx @funpixel/tools model --local normalize hero.glb --height-m 1` (unit height), `model --local decimate hero.glb --target-triangles 3000`, `model --local inspect hero.glb` (triangle count). The hosted `inspect_model`, `optimize_model` and `fracture_model` do the same on a model a job made. Joining meshes and moving the origin to the base centre are not in the bench; do them in Blender by hand or in your own script.
 - Proof render: the hosted `render_model`, or `npx @funpixel/tools model --local render hero.glb --frames 3`, gives the three angles.
 - Split a GLB for Unity with gltf-transform (`gltf-transform copy model.glb model.gltf` writes `.gltf` + `.bin` + texture files); name the textures by role (`model__basecolor`, `model__normal`, `model__mr`) so the import rule can find them. gltf-transform also decodes web GLBs (meshopt + WebP).
-- Name split textures by role and keep them under `Art/` or `Models/`, so `TextureBudget` finds them; a game without the package writes the same rule as an `AssetPostprocessor` in `Assets/Editor/`, or uses the one `gen --unity` writes.
+- Name split textures by role and keep them under `Art/` or `Models/`, so the import rule finds them: write it as an `AssetPostprocessor` in `Assets/Editor/`, or use the one `gen --unity` writes.
 - Collision and size audits are your own Edit Mode tests until the checks ship: compare each prop's `Renderer.bounds` with its `Collider.bounds`, and read the Editor's build report for texture formats.
 
 Checks: `texture_audit`, `drawn_vs_solid`, `glb_import`, `tri_budget` and `lfs_audit` are proposed, not built (partial versions exist in single games). Until then `model --local inspect` gives the triangle count.
@@ -45,7 +45,7 @@ Checks: `texture_audit`, `drawn_vs_solid`, `glb_import`, `tri_budget` and `lfs_a
    | Kind | Setting |
    |---|---|
    | Model textures | ASTC 6×6, mipmaps, sRGB only for basecolor and emissive, max 512 (props) or 1024 (stalls) |
-   | 3D models | default glTFast import of the split `.gltf`; no LODs except The Hot Gates' mid LOD |
+   | 3D models | default glTFast import of the split `.gltf`; no LODs except one game's mid LOD |
    | Rigged characters (FBX) | Generic rig, root `Hips`, Loop Time on, Apply Root Motion off, in-place clips at 30 fps |
 
 8. Confirm the import (`glb_import`): every GLB yields at least 1 mesh with more than 0 vertices, and no `.webp` or `.woff2` sits under Assets.
@@ -55,24 +55,24 @@ Checks: `texture_audit`, `drawn_vs_solid`, `glb_import`, `tri_budget` and `lfs_a
 12. Audit the build (`texture_audit`): 0 unwhitelisted RGBA32 textures, RGBA32 total under 1 MB. Mipmapped textures must be power of two; Unity 6000.6 ships a mipmapped non-power-of-two texture as RGBA32 even with an ASTC override.
 13. Store baked binary data as `.bytes` in LFS (`lfs_audit`: 0 tracked binaries over 500 KB outside LFS, no untracked paid art).
 
-Rigs: one shared skeleton gives one Animator for the whole cast. Large crowds use no rigs at runtime: The Hot Gates bakes rigged GLBs to vertex-animation textures.
+Rigs: one shared skeleton gives one Animator for the whole cast. Large crowds use no rigs at runtime: one game bakes rigged GLBs to vertex-animation textures.
 
 ## Worked example
 
-Noodle Nights (a new 3D idle game), done with the studio's own scripts (proposed for the kit). Steps 2–4 are the report's general pipeline; it does not log them per game.
+A new 3D idle game, done with the studio's own scripts (proposed for the kit). Steps 2–4 are the report's general pipeline; it does not log them per game.
 
 1. Props and stalls as generated GLB; characters as Meshy-rigged FBX (the auto-rig survived import 7 of 7).
 2. Blender prep at 3,000 triangles: joined, unit height, origin at base centre.
 3. A 3-angle proof render and the triangle count.
 4. Split: `.gltf` + `.bin` + `model__basecolor`, `model__normal`, `model__mr` PNGs.
 5. The texture import rule: ASTC 6×6, max 512 for props and 1024 for stalls.
-6. Result: data 315 → 61 MB; the app 386 → 106 MB with the build size settings also on. Porcelain Panic went 229 → 75 MB.
+6. Result: data 315 → 61 MB; the app 386 → 106 MB with the build size settings also on. Another game went 229 → 75 MB.
 
 ## Common mistakes
 
-- Importing a web GLB as it is. Dawg Ball's 12 heads "did not import at all", with no error.
-- A prop with no collider, or art bigger than its collider. 42 Pocket Crew props could be walked through; Dead Letter's Paperweight art was 54 px taller than its collider; a Dawg Ball head drawn 1.15× its hitbox was removed.
-- `shader_feature` on a runtime-created material. Pocket Crew's instanced sets drew at the origin on device.
-- Baked data as serialised YAML in git. The Hot Gates' vertex-animation crowd assets are about 70 MB of YAML (minotaur 13.0 MB, cyclops 11.8 MB), and every re-bake adds about 70 MB to history.
+- Importing a web GLB as it is. One game's 12 heads "did not import at all", with no error.
+- A prop with no collider, or art bigger than its collider. 42 props in one game could be walked through; another's paperweight art was 54 px taller than its collider; a head drawn 1.15× its hitbox was removed.
+- `shader_feature` on a runtime-created material. One game's instanced sets drew at the origin on device.
+- Baked data as serialised YAML in git. One game's vertex-animation crowd assets are about 70 MB of YAML (minotaur 13.0 MB, cyclops 11.8 MB), and every re-bake adds about 70 MB to history.
 - Paid generations outside LFS: about 840 MB lived on one Mac.
-- Trusting Meshy's Idle preset. It bowed the head to the chest (Noodle Nights) and was re-keyed in Blender.
+- Trusting Meshy's Idle preset. It bowed the head to the chest and was re-keyed in Blender.

package/plugins/claude/skills/art/references/animation.md

@@ -38,7 +38,7 @@ is edited from, so the frames stay that character.
   the distinct poses per cycle.
 - The frames share one scale and one ground line; the feet are each frame's anchor.
 - The sheet draws every frame from the side or the front. An actor seen from above (a maze
-  chase, a twin-stick arena) has no ground line, and a Riad Rush run got a side-view gecko back
+  chase, a twin-stick arena) has no ground line, and one run got a side-view gecko back
   for a top-down maze (57 credits, discarded). Give such an actor its two to four key poses as
   `edit_image` pose edits of the top-down sprite (below), stepped at the stride rate.
 - Measured (the studio's animation evals): the character holds across frames, but each

package/plugins/claude/skills/art/references/cast.md

@@ -1,7 +1,7 @@
 # The cast: lock first, check second
 
-A game with a cast draws each member dozens of times: Gallery Heist drew its keepers 6 to 13 times
-each and its crew in rest, brace and cover poses; Held Note kept front, back and left pictures of
+A game with a cast draws each member dozens of times: one heist game drew its keepers 6 to 13 times
+each and its crew in rest, brace and cover poses; another kept front, back and left pictures of
 every character. What kept them on model was never a better sentence. It was the same picture
 attached to every call, and a look at each result against that picture before it was kept.
 
@@ -68,6 +68,6 @@ measure on any picture.
 ## What the numbers cannot see
 
 They catch a recolour, a lost signature colour, the wrong subject and a changed build. Two crew
-members in one uniform read close by numbers (Gallery Heist's Smudger and Carrier: 58% outline
+members in one uniform read close by numbers (one heist game's two crew members: 58% outline
 overlap, the same colours). Named features (the face, the earring, what the hand holds) are the
 identity seat's: `npx @funpixel/tools judge identity --concept design/cast/ren/master.png <views…>`.

package/plugins/claude/skills/art/references/effects.md

@@ -57,7 +57,7 @@ what each one measured. Pick what serves the look the game chose; a clean flat s
 9. **In a whole 3D game, read the slow frames.** Averages hide hitches: read p95/p99 at a named
    beat (the first fight, not the whole run), and name what appeared on each slow frame (a shader
    compile, a texture upload) before fixing it. Warm-up, culling skinned actors, a pixel budget,
-   pacing to the refresh, and the leaks Held Note met: [frames.md](frames.md).
+   pacing to the refresh, and the leaks one game met: [frames.md](frames.md).
 
 ## Worked example
 
@@ -76,7 +76,7 @@ textures; at resolution 1 it held 4.1 MB, and on the software renderer its frame
 - A custom Pixi filter that redeclares `uInputSize` in the fragment shader: the precisions differ
   from the vertex stage and the program fails to link (a black screen). Pass a varying instead.
 - `EffectComposer` without `OutputPass`: colours come out dark and tone mapping is skipped.
-- A white starburst or full-screen flash on contact: Particle Rush's (Oct 2026) read as glare, one 38% clipped.
+- A white starburst or full-screen flash on contact: one racer's (Oct 2026) read as glare, one 38% clipped.
   A small warm glow at the contact point, about 0.16 s, answers the hit.
 - Dither or halftone transparency in a shader: at phone size it reads as a glitch. Fade in broad, slow bands or
   real alpha. To find which module owns a fault, capture with one module off

package/plugins/claude/skills/art/references/frames.md

@@ -2,13 +2,13 @@
 
 The recipe lab ([cost](cost.md)) measures one effect in a small scene. A whole game adds things a
 lab never meets: first-use compiles, a crowd of skinned actors, level rebuilds, big retina windows
-and displays faster than 60 Hz. This is what Held Note (a three.js third-person horror game, 29 Sep
+and displays faster than 60 Hz. This is what one game (a three.js third-person horror game, 29 Sep
 2026, `design/PERF.md` in its repo) measured, and what each fix bought. The numbers are that game's
 on that machine; the causes recur.
 
 ## Smooth is a distribution, not an average
 
-At 78 fps average Held Note still stuttered. Read p95, p99 and the worst frame, and where the slow
+At 78 fps average that game still stuttered. Read p95, p99 and the worst frame, and where the slow
 frames land. Then name each slow frame's cause before changing anything: on a hitch frame, read
 `renderer.info` for what appeared (new shader programs, textures, geometries). Every hitch it
 found had a cause that showed there.
@@ -21,7 +21,7 @@ Profile vsync-capped, as a player sees it. An uncapped run (`--disable-frame-rat
 the game's own refresh detection into cutting resolution at 2.5 ms frames. And profile on a machine
 nobody is playing on: films and profilers share the GPU and read as lag to the player.
 
-## What Held Note found, and what each fix bought
+## What that game found, and what each fix bought
 
 | Symptom | Cause (measured) | Fix |
 |---|---|---|
@@ -44,13 +44,13 @@ Results on the same recorded run: desktop 1600×900 @2x worst frame 87 → 29 ms
 A 1950×1090 window at pixel ratio 1.5 is 4.8 M pixels through AO, depth of field and bloom:
 GPU-bound (p99 26 ms). On a 120 Hz display, 9 ms frames flip between 8 and 16 ms and judder.
 
-- **A pixel budget** caps the ratio by pixel count (Held Note: 2.6 M on desktop, 1.3 M on a phone);
+- **A pixel budget** caps the ratio by pixel count (that game: 2.6 M on desktop, 1.3 M on a phone);
   the DOM HUD stays at full retina.
 - **Pace to the refresh.** Detect the refresh from the fastest frames, judge each second's p90
   against that slot, and when 120 cannot hold, lock to 60 (render every other refresh) before
   lowering resolution.
 - **Hysteresis.** Each resolution change rebuilds the post-processing targets (41 textures, a 70–100
-  ms frame), so an oscillating scaler stutters every few seconds. Held Note's rule: no step up within
+  ms frame), so an oscillating scaler stutters every few seconds. Its rule: no step up within
   12 s of a step down, five fast seconds to earn one, none after two downs.
 
 Result: 119.2 fps on a 120 Hz display at 1950×1090 @2x, p99 9.4 ms, worst 17.5 ms, zero hitches.
@@ -74,7 +74,7 @@ with its animations paused, removed it.
 
 ## When something glows that should not
 
-"Anything that looks like a white ball is a light" held four times in Held Note: the torch's
+"Anything that looks like a white ball is a light" held four times in that game: the torch's
 hotspot on a close surface, the beam cone seen end-on, coral pushed past the bloom's white point,
 and the worst one, an additive ray-marched beam seen down its own axis. From the camera behind her,
 the view ray ran down the whole lit cone and eight summed samples bloomed into a white disc that

package/plugins/claude/skills/art/references/judging-frames.md

@@ -1,6 +1,6 @@
 # Judging frames against the targets: a fixed rubric, fixes by module, eye comfort
 
-From Particle Rush (Oct 2026), a Unity 6 3D racer built by several module agents (world, racers, levels) and one
+From a Unity 6 3D racer (Oct 2026) built by several module agents (world, racers, levels) and one
 integrator. An independent judge scored real captures against the painted targets every round; what follows is
 what held up, with its numbers. `funtools score` measures the same frames (layout, palette and light, density,
 comfort); this is the judge that reads what numbers cannot.
@@ -16,7 +16,7 @@ comfort); this is the judge that reads what numbers cannot.
 
 - **The same rubric every round**, written before seeing the frames, so a score moves only when a frame changed.
 - **Full-resolution phone captures from the real build** (the player on the target device size, not the editor, not
-  a web stand-in): Particle Rush judged 1172×2532 frames; half-size shots hid the material detail being scored.
+  a web stand-in): the racer judged 1172×2532 frames; half-size shots hid the material detail being scored.
 - **Per frame:** a total, the four category scores, the change since last round, and one eye-comfort word
   (yes, mild, glare, busy, no): could a person look at it for a minute without strain?
 
@@ -26,7 +26,7 @@ Every fix names the module that owns it, `[WORLD]`, `[RACERS]`, `[LEVELS]` (or t
 frame, the pixel region and the number: "[WORLD] the sun bloom washes out the upper right (11% clipped): bloom
 intensity −50%, exposure −0.4 EV". The integrator routes each tag to that module's agent (its lane, where
 `funtools lanes` runs them), so no agent reads another's list. End with the five changes that would raise the average most,
-each with its expected gain; Particle Rush's round-2 list predicted 76–80 from 64.
+each with its expected gain; the racer's round-2 list predicted 76–80 from 64.
 
 **Find the owner before assigning a fault.** Capture the same frame with one module turned off (a launch flag
 per module: no world dressing, no racer effects). The fault that disappears belongs to the module that was off;
@@ -34,7 +34,7 @@ a stripe under the racer that stays with the effects off is not the effects agen
 
 ## Read the categories, not only the total
 
-Particle Rush's round 2 rose 55.6 → 64.2, but composition gave +5.1 and effects +2.2 while light gave +0.7 and
+The racer's round 2 rose 55.6 → 64.2, but composition gave +5.1 and effects +2.2 while light gave +0.7 and
 detail +0.6: the gain came from removing noise (a white halo, crossing speed lines, a fourth trail), not from
 adding quality. Calm is not finished. When light and detail stall, give materials and lighting a round of their
 own (per place: a key light, a rim, a regrade, authored hero surfaces, a bake-off of each hero asset against
@@ -48,17 +48,17 @@ frames than the targets are the goal, as long as the quality matches.
 - `funtools score` prints each frame's **busyness** (the mean brightness step between neighbouring pixels on a grey
   copy 360 px on its short side) and **clipped share** (pixels at luma ≤ 2 or ≥ 253) beside its target's, and
   stamps the zero board's `comfort` counter. It warns, never fails: over 14, over 1.25× the targets' busiest, or
-  over 6% clipped. Particle Rush's painted targets read 9.2–15.4 there and its web frames 6.9–8.5; our liked
-  games' frames 2.0–11.9; Gallery Heist's room read 7.8% clipped before its light round and 0 after.
+  over 6% clipped. The racer's painted targets read 9.2–15.4 there and its web frames 6.9–8.5; our liked
+  games' frames 2.0–11.9; one heist game's room read 7.8% clipped before its light round and 0 after.
 - A calm frame has room for detail: add mid-ground surface detail at about half the target's density, low
-  contrast and slow (Particle Rush's near road, on its judge's own scale: a bottom-quarter gradient of about 3,
+  contrast and slow (the racer's near road, on its judge's own scale: a bottom-quarter gradient of about 3,
   against 1–2 then and 5–6 in the targets).
 - Clip sources are fixes, not taste: blown bloom, a white flash, an emissive that clips to white instead of
   keeping its colour (clamp it).
 
 ## Framing calls are the owner's
 
-A rubric's framing numbers are a judge's proposal. Particle Rush's rubric asked for the racer at 80% of the frame
+A rubric's framing numbers are a judge's proposal. The racer's rubric asked for the racer at 80% of the frame
 width; the owner settled on about 40%. Record the owner's number as a rule in the GDD (a tuning row), tell the
 judge it is fixed and not to mark it down, and judge what fills the space the smaller hero leaves instead. Ask
 the owner before a judge's hero size, camera height or crop becomes a fix.

package/plugins/claude/skills/art/references/look.md

@@ -23,7 +23,7 @@ game, the moves below are what held our bar games' looks together.
    procedural proven by a rendered study; "primitives for now" is a stand-in, not a route).
    At the first look the world's kit stands beside the hero: sheets of its structures, ground,
    props and living cast, cut (`npx @funpixel/tools cut`) before the slice is built from them;
-   without it a fresh Blue Freight build coded its world from shapes.
+   without it one fresh build coded its world from shapes.
 2. **Write five AD rules.** The critic panel reads them:
    each checkable in a frame (colour roles, outline weight, light, material, actor readability at
    a stated phone size); no universal percentage for hero size.
@@ -36,7 +36,7 @@ game, the moves below are what held our bar games' looks together.
    paying for new art. Check phone and desktop; a rule revised to excuse a defect stops measuring anything.
    For a game with a moment ladder, a target frame per moment and captures of the same seconds
    from a recorded run make a loop: name the biggest difference, fix it, recapture. Judge
-   generated models under the game's own light, where Held Note's near-black textures were
+   generated models under the game's own light, where one game's near-black textures were
    lifted into white blow-outs. Both, with what they found:
    [prompts.md](prompts.md#target-frames-at-moments-and-the-loop-that-closes-the-gap).
    Match the targets' quality, not their density (`score` warns on a busy or clipped frame); the rubric,
@@ -55,7 +55,7 @@ game, the moves below are what held our bar games' looks together.
   a target, hero or key art where the comparison is worth its cost.
 - Before the game exists, `npx @funpixel/tools look --sheet` puts every sprite at play heights
   (24, 48, 96 px) on a neutral band and on the real ground colour (`roles.ground` or `--ground`)
-  and measures each against it. The Hot Gates' sheet found its terracotta army vanishing on the
+  and measures each against it. One game's sheet found its terracotta army vanishing on the
   pass floor at 24 px before any game code.
 - A pixel-art look can be exact rather than imitated: `pixelArt` (scale, palette) in
   `funpixel.json` has `assets process` and `gen` snap every picture to the grid and palette, and
@@ -89,7 +89,7 @@ funpixel.json `art3d` declares it with its reason and a proof frame (skill `fp:3
 
 ## Worked example
 
-Fold Room's sandstone rooms wore dark pills: `look --hud` read system type, 20 dark plates and
+One game's sandstone rooms wore dark pills: `look --hud` read system type, 20 dark plates and
 the scaffold's glyphs. A `## UI` section chose sandstone plaques, a carved face and an inked icon
 family; a generated plate went in through `border-image`.
 

package/plugins/claude/skills/art/references/pipeline.md

@@ -2,7 +2,7 @@
 
 When a game makes more than a handful of assets from text, the games that shipped best kept them
 as data: what each piece is, one style file, and a record of what made each file. Four of our
-games (The Last Good Robot, The Hot Gates and two more) arrived at the same shape on their own.
+games arrived at the same shape on their own.
 `funtools assets` is that shape, on the shelf. Use all of it, part of it, or none: hand-made,
 procedural and one-off `gen` assets need no manifest, and nothing checks for one.
 
@@ -78,7 +78,7 @@ changed subject; `gen --only <name>` tries the same words again.
 
 ## At scale: the asset script
 
-Our own games carry 70 to 250 assets (Gallery Heist: 116 paintings, 183 sounds, 11 music cues, 25
+Our own games carry 70 to 250 assets (one heist game: 116 paintings, 183 sounds, 11 music cues, 25
 models). Nobody calls a tool 200 times by hand. The plan's `## Asset script` is the list, one row
 per family with a count, and it becomes the manifest:
 

package/plugins/claude/skills/art/references/prompts.md

@@ -119,7 +119,7 @@ Taste goes to the `critic` agent, which grades craft within the declared style.
 
 ## Target frames at moments, and the loop that closes the gap
 
-Held Note (a three.js horror game, 29 Sep 2026) made its target frames for moments of the GDD's
+A three.js horror game (29 Sep 2026) made its target frames for moments of the GDD's
 moment ladder (0:10 the moon pool, 1:00 the first enemy, 2:00 the dodge, 3:00 the splice, 5:00 the
 seam), so each capture has one picture it should match. The loop that closed its look gap:
 
@@ -134,13 +134,13 @@ floor fixed it), a camera inside lockers the hero could walk into (collision), a
 inside a door. The owner's note that started the second art pass was the same comparison: "what
 we imagined was better; we must get close to it".
 
-A capture only counts if it shows the moment. Held Note's third panel scored 4.9 because every
+A capture only counts if it shows the moment. That game's third panel scored 4.9 because every
 view silently sat at t = 0 behind the opening shot after a refactor; looking at the captures before
 reading the score found it.
 
 ## Judging generated art in the game's own light
 
-A generator's preview is lit for the generator. Under the game's light Held Note's generated
+A generator's preview is lit for the generator. Under its own light that game's generated
 textures read near-black, so the enemy's albedo was lifted ×1.9 to read in the dark; a few metres
 into the torch's spot it then went pure white with no face. Albedo above 1 has nowhere to go under
 a strong light. The lift went back to 1.05 and a cold rim light did the reading-in-the-dark job
@@ -153,7 +153,7 @@ When something white appears that should not, suspect a light first and hide one
 
 ## A rubric for a 3D cast, written before generating
 
-Held Note wrote ten lines before its first 3D model, each tied to an AD rule or a target frame, and
+The horror game wrote ten lines before its first 3D model, each tied to an AD rule or a target frame, and
 turned every model on a turntable against them. Three show the shape: "light is the palette:
 surfaces desaturated, the only saturated hues amber (station), cyan (hero, safe tech), coral
 (enemies)"; "reads at phone size: at about 120 px tall the silhouette alone says what it is";

package/plugins/claude/skills/art/references/ui.md

@@ -5,17 +5,17 @@ in the game's world, played on a real phone and judged every round with the same
 characters and the levels, never a layer added at the end. The controls count too: the gesture
 on the play, how it feels and how it answers, is part of the look.
 
-The owner played our games on an iPhone (30 Sep 2026) and named the same fault in each. Fold
-Room's warm sandstone rooms wore a HUD of dark app pills with cyan outlines and system type, and
-its title was a white bold wordmark over the scene with a cyan pill and three icon circles. Floe's
-frosty board sat under the same pills and stock glyphs ("better icons, not emojis"). Saltwake had
-heavy navy slabs; Plush Pop's title stacked panels over its basket; Lantern Night's title floated a
-small dark board in an empty sky. Riad Rush looked better only because its creator happened to
-carve its HUD from the riad's wood. The cause was ours: the look's direction covered the world,
+The owner played our games on an iPhone (30 Sep 2026) and named the same fault in each. A puzzle
+game's warm sandstone rooms wore a HUD of dark app pills with cyan outlines and system type, and
+its title was a white bold wordmark over the scene with a cyan pill and three icon circles. An ice puzzle's
+frosty board sat under the same pills and stock glyphs ("better icons, not emojis"). A racer had
+heavy navy slabs; a toy game's title stacked panels over its basket; a night game's title floated a
+small dark board in an empty sky. One game looked better only because its creator happened to
+carve its HUD from its world's wood. The cause was ours: the look's direction covered the world,
 the cast and the target frames, but not the interface, so every game fell back to the scaffold's
-dark pill. All three of Fold Room, Floe and Lantern Night still ship the scaffold's own glyphs.
+dark pill. Three of them still ship the scaffold's own glyphs.
 
-We had the answer already. The Last Good Robot's GDD: "Every screen is drawn in the factory's own
+We had the answer already. One of our games' GDD: "Every screen is drawn in the factory's own
 kit (riveted steel plates, one hazard stripe, stencil type) and speaks in the world's voice. A
 screen that isn't specified gets invented by whoever reaches it first." So the look covers the UI
 by default, and the GDD specifies every screen in the world's kit: the screen list (title, how to
@@ -58,7 +58,7 @@ plan-at-size): a plate, a button in its states, a frame for cards, the icon fami
    with the game (two families at most); a system stack is a choice to name, not a default.
 3. **One icon family, drawn like the art.** Pause, sound, hint, undo, stars and the currency are
    one set: the art's outline weight, its corner radius, its two-tone fill, the palette's roles, a
-   motif of the world (Fold Room's sun banner, Floe's snowflake). Never a stock glyph set or emoji.
+   motif of the world (a sun banner, a snowflake). Never a stock glyph set or emoji.
    Read them at 24 and 48 px on the plate they sit on.
 4. **Colour roles from the palette.** Plate, plate edge, ink on the plate, the one accent (and what
    it means), harm. Contrast is computed: 4.5:1 for text, 3:1 for icons (skill
@@ -106,7 +106,7 @@ better. Five that fit most genres:
 | Objects of the world | plates, plaques, tags and buttons made of the world's material | puzzle, adventure, cosy, 3D rooms |
 | The frame holds the numbers | the board's or jar's own frame carries score and next | merge, match, board puzzles |
 | The thin bar | one slim strip along an edge, everything else on the actors | cards, strategy, roguelikes |
-| On the actor | health and state on the hero or the tool (Held Note: "the suit is the HUD", a wrist gauge strapped to her arm) | action, platformers, 3D |
+| On the actor | health and state on the hero or the tool (a horror game: "the suit is the HUD", a wrist gauge strapped to her arm) | action, platformers, 3D |
 | Paper and ink | margin notes, stamps, torn labels, a hand-drawn line | narrative, cosy, crafting |
 
 Then make it this game's: `npx @funpixel/tools variety` sets the title screen beside the other
@@ -133,8 +133,8 @@ game's style from its style frame and saved under `design/first-look/`; each is
 the plan and judged like the HUD:
 - **The title** is the game's first promise and the frame a player judges first. Title key art:
   the game's own scene, never a board preview floating on a flat colour. The overlay comes from the
-  world: the wordmark carved into or lit on its surface, the start as a thing of the world (Fold
-  Room's glowing fold: step in to start), menu buttons of the world's material, icons drawn like
+  world: the wordmark carved into or lit on its surface, the start as a thing of the world (a
+  glowing fold: step in to start), menu buttons of the world's material, icons drawn like
   its motifs, the caption in the display face. Ask of it what a player asks: does it make me want
   to press Play? A white bold wordmark over a pill and three icon circles is the overlay the
   owner called plain, however good the scene behind it.
@@ -150,8 +150,8 @@ the plan and judged like the HUD:
   `fp:verifying`, fidelity).
 - `npx @funpixel/tools look --hud` reads the play and the title for the scaffold's clothes: the
   system type stack, dark neutral plates, the scaffold's own glyphs, icons typed as characters (a
-  "★" or an emoji in a label, what the owner called "emojis" on Floe), a wordmark that is plain
-  system text (a gradient washed over it is still system text: Floe's R4 title). It informs; a creator who chose any of them names why in the hand-back.
+  "★" or an emoji in a label, what the owner called "emojis" on one game), a wordmark that is plain
+  system text (a gradient washed over it is still system text: one game's R4 title). It informs; a creator who chose any of them names why in the hand-back.
 
 How to dress the scaffold's HUD and screens, with worked examples: skill `fp:mobile-ui`, its
 reference `restyle-ui.md`.

package/plugins/claude/skills/art/references/unity.md

@@ -4,7 +4,7 @@
 
 ## Tools
 
-The studio's `com.joydle.core` package carries two of the import rules (fp:unity-game, `references/studio-packages.md`): `SplashImport` scales the studio splash's parts to a power of two with mipmaps, and `BuildTools.TestFlight` refuses to build without `Assets/_Game/Art/Icon/app_icon.png`, importing it at 1024, uncompressed, no alpha. Not shipped by the kit or the package: the Unity starter, its checks and the studio scripts (`key_green.py`, `cut_sheet.py`, a font builder). Do the rest with what exists:
+Have the build method refuse to build without `Assets/_Game/Art/Icon/app_icon.png`, imported at 1024, uncompressed, no alpha. Not shipped by the kit: the Unity starter, its checks and the studio scripts (`key_green.py`, `cut_sheet.py`, a font builder). Do the rest with what exists:
 
 - Generate with the hosted `generate_image` or `generate_sprite`, or `npx @funpixel/tools gen sprite "<prompt>" --size 1024x1024 --references target.png --unity .`: into `Assets/_Game/Generated/Sprites/` (`Images/` for plates, `Textures/` for tiles), imported on first sight as Sprite, no mipmaps, ASTC 5×5 (plates 6×6), by the rule it writes beside them (`Generated/Editor/FunpixelGenImport.cs`; no `.meta` written). Sign in once with `npx @funpixel/tools login`. Give the size: a 9:16 default made a 720×1280 icon. The studio's `gen_image.py` calls the provider with a `.env` key: no metering, no cost log.
 - Key with the hosted `key_sprite`, or ask for `--transparent` art; cut a sheet with `npx @funpixel/tools parts sheet.png --grid 4x2 --key auto` (or `cut` for a painted reference sheet).
@@ -18,7 +18,7 @@ Checks: `texture_audit`, `glyph_coverage`, `tmp_assets_small` and `import_clean`
 ## Procedure
 
 1. Generate sheets against the game's target image, on a green screen, and key them (`key_sprite`, or `--transparent` art). Background removal with BiRefNet ate white objects (a whole plate).
-2. Read every word in generated art by OCR; the report's example is "JOTDLB". Which OCR tool is not in the report.
+2. Read every word in generated art by OCR; a misspelt logo got through. Which OCR tool is not in the report.
 3. Check sprite fill: 0.85 or more of the canvas. One sprite filled only 73%. How to measure fill is not in the report.
 4. Rasterise at the size the sprite is drawn. The reason is not in the report.
 5. Reuse web art: `dwebp` to PNG, pivots from the web's anchors at 100 pixels per unit.
@@ -31,7 +31,7 @@ Checks: `texture_audit`, `glyph_coverage`, `tmp_assets_small` and `import_clean`
    | Splash, any mipmapped art | scaled to power of two |
    | App icon | 1024×1024, uncompressed, no alpha |
 
-7. Pick the ASTC block by role and measure it: round-trip each atlas with astcenc, and use 4×4 if it scores under 35 dB. 6×6 put ringing on Scrap Riot's ink outlines (27–35 dB); 4×4 everywhere costs +2.4 MB.
+7. Pick the ASTC block by role and measure it: round-trip each atlas with astcenc, and use 4×4 if it scores under 35 dB. 6×6 put ringing on one game's ink outlines (27–35 dB); 4×4 everywhere costs +2.4 MB.
 8. Use one material per atlas.
 9. Match colour space to the source renderer on day one. Pixi and canvas blend in gamma; under Linear their translucent overlays were about 2× too strong.
 10. After swapping a PNG, check its sprite rect. A PNG of a new size keeps the old rect and Unity drops the sprite; re-importing an atlas regenerates sprite ids. `import_clean`: a fresh import gives 0 `.meta` diffs and every rect fits its texture.
@@ -42,37 +42,37 @@ Text and glyphs:
 12. Import TMP Essential Resources first, once in the Editor, or `CreateFontAsset` returns null (`funtools doctor` fails a project without them).
 13. Build fonts from the full font file. A web subset lacked Ş, Ğ and İ; Nunito lacks ▶, ◀ and ★.
 14. Run `glyph_coverage`: every character in UI strings and names exists in the atlas, 0 missing.
-15. Warm every glyph behind the title screen. First-use baking cost 183–266 ms in Dawg Ball and 50–270 ms in Pocket Crew (17–20 ms after warm-up); a missing glyph was The Hot Gates' last 48 ms hitch.
+15. Warm every glyph behind the title screen. First-use baking cost 183–266 ms in one game and 50–270 ms in another (17–20 ms after warm-up); a missing glyph was a third game's last 48 ms hitch.
 16. Atlases are 1024×1024 SDF, sampling 64–90, padding 8–12. The report does not settle static against dynamic.
 
    | Atlas | Cost | Seen in |
    |---|---|---|
-   | Static | about 5 MB for 1024² atlases | Scrap Riot, which then went dynamic |
-   | Dynamic | rewritten by every batch run, conflicts on every merge, 8 MB each in git after captures | Pocket Crew (`Nunito ExtraBold SDF.asset`) |
+   | Static | about 5 MB for 1024² atlases | one game, which then went dynamic |
+   | Dynamic | rewritten by every batch run, conflicts on every merge, 8 MB each in git after captures | one game (`Nunito ExtraBold SDF.asset`) |
 
    If dynamic, clear the atlases after captures. `tmp_assets_small`: every `*SDF.asset` in git under 50 KB.
 17. Ship TMP outline and underlay variants as material assets (`Resources/UiFontVariants`). Variants of materials made at runtime are stripped from release builds.
 18. Use `textWrappingMode`; `enableWordWrapping` is obsolete in Unity 6 (`funtools doctor` warns).
-19. Judge icon candidates small and against competitors: The Hot Gates masked them at 100, 60 and 40 px in a mock search row with 41 competitor icons, for $5.95.
+19. Judge icon candidates small and against competitors: one game masked them at 100, 60 and 40 px in a mock search row with 41 competitor icons, for $5.95.
 
 ## Worked example
 
-Dead Letter (a 2D port from Pixi), texture pass:
+A 2D port from Pixi, texture pass:
 
 1. Web sprites as PNG; pivots from the web's anchors at 100 pixels per unit, set in the game's setup script.
 2. Colour space switched from Linear to Gamma, because translucent overlays were about 2× too strong.
 3. ASTC by role: painted plates 8×8, sprites 6×6, hero and HUD art 5×5.
 4. Result: textures 69 → 50 MB with no visible change.
-5. The power-of-two splash fix (PR #14) made every game 7–8 MiB smaller; Koi Out's resources went 11.7 → 3.8 MB.
+5. The power-of-two splash fix made every game 7–8 MiB smaller; one game's resources went 11.7 → 3.8 MB.
 
 ## Common mistakes
 
 - A mipmapped texture that is not a power of two. The splash shipped 10.3 MB raw in every game.
-- One block size for everything. 6×6 rang the ink outlines on Scrap Riot's atlases.
-- Text in generated art not read: "JOTDLB" (game not named).
+- One block size for everything. 6×6 rang the ink outlines on one game's atlases.
+- Text in generated art not read: a misspelt logo (game not named).
 - Art that does not fill its canvas: spool art at 73% fill (game not named).
 - A swapped PNG that kept its old sprite rect: an empty logo on TestFlight (game not named).
 - A subset font: "KARABAŞ" broke (game not named).
-- Outline variants not shipped: outlined numbers drew as cream blobs in The Hot Gates' player build.
-- Dynamic atlases left after captures: 8 MB each in git. Two review builds (Dead Letter, Scrap Riot) carried uncommitted reimports of 4 TMP font assets, so main no longer matched what Apple reviewed.
-- Icons by prompt: Dawg Ball spent $21.46 over 7 rounds before a pick; round one came out 720×1280 from the 9:16 default.
+- Outline variants not shipped: outlined numbers drew as cream blobs in one game's player build.
+- Dynamic atlases left after captures: 8 MB each in git. Two review builds carried uncommitted reimports of 4 TMP font assets, so main no longer matched what Apple reviewed.
+- Icons by prompt: one game spent $21.46 over 7 rounds before a pick; round one came out 720×1280 from the 9:16 default.