All skills
thedivergentai avatar

/godot-agent-vision

@90634df

Agentic eyes for Godot work: cross-platform screenshot CLI (asset/region/window/screen) to budgeted WebP, TEMP editor viewport bridge under .gdskills/, and Taste Receptor Atlas v2 (~237 micro-receptors × 0–2: hierarchy, contrast, typography×41, affordance, space, color, icon, composition, light/FX, motion-tells, anti-slop, originality, identity). Use when agents must see UI, assets, or the editor — not for in-game screenshot Autoloads. Keywords: agent vision, screenshot, WebP, visual QA, taste receptors, typography, UI review, asset sheet, editor viewport, anti-slop.

Use this Skill: https://skilld.dev/gh/thedivergentai/gd-agentic-skills/godot-agent-vision

This session only. Nothing lands on disk.

referencesidentity-sight.md

≈1.2k tokens on demand. Your agent reads this file only when SKILL.md points to it.

Graphic identity sight (system from pixels)

Deep load for Family ID (+ COLOR/ICON bridge) in taste-receptors.md (rollup V8). Cohesion is what makes simple assets feel premium. Identity failure = one frame looks like a different game.

Map every fail to an ID-* / COLOR-* / ICON-* receptor ID in the zero list.

Do not reward purple-indigo mesh gradients, iridescent blobs, or Lucide-in-rounded-squares as identity unless clearly owned by world rules and repeated with discipline.


Anti-patterns

Tell Screenshot evidence Damage
Default gradient identity Purple/indigo/pink or holographic mesh as only “brand” Interchangeable AI web template
Palette infinity Unrelated saturated hues with equal authority No learnable danger/reward language
Shape language drift Soft rounded UI + razor hard-surface world (no reason) Kitbash product
Line-weight lottery Outlines thick on characters, absent on props, sketchy on UI No style guide
Render technique clash Flat cel + PBR prop + painterly BG Asset-pack collage
Lighting direction chaos Each sprite private light rig World not singular
Typeface tourism Pixel HUD + script title + grotesque body + blackletter captions No typographic identity
Icon metaphor chaos Mixed skeuomorphism levels/grids Chrome not a system
UI divorced from world Menu stylesheet ≠ game palette/edges Menus = other product
Marketing vs gameplay split Store key art ≠ in-engine HUD/world Brand breaks at screenshot seam
Second-wave slop identity Cream + terracotta + italic serif, no world link New AI median, still generic
Equal accent screaming Every panel uses accent Accent stops meaning emphasis

False positive: Dark UI + one glow is common in human tools. Fail only when many defaults stack and no distinctive motif/shape/type appears.


Positive criteria

Palette as a system

  • Limited family dominates (~8–16 color family feel in stylized 2D is a common production target).
  • Roles readable: neutral world, danger, reward/loot, interactable, UI emphasis.
  • UI colors from same family as world — not detached SaaS kit.

Shape language & motifs

  • Edge vocabulary repeats (rounded / chamfered / ornate / chunky / geometric).
  • Recurring motifs as a controlled set — not one-off stickers.
  • Logo/mark shares shape language with UI frames/icons.

Typography identity (pairs with V3)

  • Heading vs body pairing stable across menu and HUD crops.
  • Type genre matches world while staying legible.
  • Hierarchy steps consistent screen-to-screen — no snowflake titles every time.

Icon & chrome system

  • Shared grid, stroke, corner radius, fill/line policy.
  • Black-silhouette test would still distinguish key icons.
  • Buttons/panels share radius, border, shadow, state language.

Lighting, material, render continuity

  • Light direction/softness/ambient tint consistent in-shot.
  • One realism dial for roughness/specular/texture density.
  • Outline presence/absence matches declared technique globally.

Cross-surface identity

  • HUD, pause menu, store/key art share palette + shape + type DNA.
  • VFX accents obey accent rules — no unbound second brand.
  • Genre tone coherent without collapsing into generic AI template.

Side-by-side drift (multiple shots)

  • Adjacent screenshots match value range, edge treatment, UI spacing, accent discipline.
  • Hero assets obey the same rules as the cast.

Sub-scores → V8

Sub-ID 2 0
ID-PAL Limited family + roles Infinite unrelated hues
ID-SHAPE Repeated edge/motif grammar Per-asset silhouette lottery
ID-TYPE Stable pairing + hierarchy Font tourism
ID-ICON Shared grid/stroke/radius/states Mixed packs
ID-LIGHT One light/material policy Private rigs / technique clash
ID-BRIDGE HUD↔menu↔key art DNA Marketing≠gameplay split
ID-SPEC Owned non-default choices AI-median template identity

Gates: ID-SPEC = 0 → identity FAIL even if pretty. ID-BRIDGE = 0 when marketing+gameplay both visible → FAIL cohesion. Purple/cream/neon defaults PASS ID-SPEC only with unique shape/type/motif ownership — never gradient alone.


Sources

  • Console Creatures — style guide / art bible contents
  • Sense Central / IndieOps — cohesion, palette roles, anti-references
  • WANDR — interface carries tone; tone ≠ illegibility excuse
  • Signs of AI design / AHD SLOP_TAXONOMY — default identity centroids

Source: SKILL.md on GitHub

1 warning1mo3 checks · Risk MEDIUM
  • Gen Agent Trust Hub1mo

    The skill captures screenshots and interfaces with the Godot editor by executing system commands and modifying project configuration files. It includes a temporary editor bridge staged as a Godot plugin, which introduces risks related to command execution and dynamic module loading.

  • Socket1mo

    No alerts

  • Snyk1mo

    Risk: LOW · No issues

Signed by skilld at 90634df. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 3 weeks ago.

Activeupdated 2 months ago

README badge

README badge for thedivergentai/gd-agentic-skills/godot-agent-vision