Skip to content

A compact Agent Skill for highly simplified, rounded, subtly neo-skeuomorphic IP mascot logos.

feat: default mascot logos to two IP colors

s1dashu25ceaaba882d
diff --git a/README.md b/README.md
index 9d543a6..ef96cc7 100644
--- a/README.md
+++ b/README.md
@@ -9,7 +9,7 @@ It follows the open Agent Skills format and is designed to work with any compati
## What it enforces
- One dominant silhouette built from roughly 6–10 basic shapes
-- Two-color or three-color total palettes, including the background
+- Three semantic colors by default: two IP base colors plus one background color
- Three candidates by default: three distinct IP directions when the subject is open, or three controlled variants when the subject is specified
- A quantified restrained-color default: softened chromatic backgrounds, warm neutrals, and explicit silhouette/detail contrast targets
- Thick, rounded forms without sharp or fragile details
@@ -35,10 +35,10 @@ For a personal installation, copy `SKILL.md` into `~/.agents/skills/ip-as-logo/`
Ask your AI agent for an IP mascot logo, for example:
```text
-Create a two-color rounded ghost IP logo on a deep navy background.
+Create a rounded ghost IP logo on a deep navy background.
```
-The skill does not ask for a monochrome or multicolor choice by default. It produces one two-color logo with one IP color plus one background, and two three-color logos with two IP colors plus one background. A two-color logo uses background-colored negative space for facial marks rather than introducing a third color.
+The skill does not ask for a color-mode choice by default. Every default candidate uses three semantic colors: two IP base colors plus one background color. It no longer reserves any fraction of the candidate set for two-color logos. Closely related highlight and shade variants may be introduced around either IP base color for the ultra-light neo-skeuomorphic effect without counting as additional semantic colors. A two-color logo is generated only when the user explicitly requests it, and then uses background-colored negative space for facial marks rather than introducing a third color.
When the user already names an IP subject, all three candidates explore that subject through controlled variations. When the subject is open, the skill defaults to three genuinely different IP directions tied to different product attributes or brand promises.
@@ -46,7 +46,7 @@ If the skill runs inside a product repository, it inspects relevant read-only co
Compatible agents may generate the three candidates in parallel with subagents. Codex can use ImageGen when available; other agent environments may use any configured image generator. If no generator is available, the skill asks the user to provide or enable one instead of pretending that an image was generated.
-When the user does not supply a palette, the skill favors clearly chromatic but restrained backgrounds rather than neon color or muddy gray. It uses OKLCH target bands when numeric control is available, prefers warm off-white with charcoal or deep navy, and keeps the normal design to no more than three semantic colors including the background.
+When the user does not supply a palette, the skill favors clearly chromatic but restrained backgrounds rather than neon color or muddy gray. It uses OKLCH target bands when numeric control is available, prefers warm off-white with charcoal or deep navy, and keeps the normal design to exactly three semantic colors: two IP base colors plus the background. Ultra-light highlight and shade variants must remain close to their corresponding IP base color.
## Repository structure
diff --git a/SKILL.md b/SKILL.md
index 233f4f2..b2a14a4 100644
--- a/SKILL.md
+++ b/SKILL.md
@@ -1,6 +1,6 @@
---
name: ip-as-logo
-description: Generate highly simplified personified IP mascot logos with Flat-first geometry, rounded heavy forms, and extremely subtle neo-skeuomorphic shading. Use when creating an animal, creature, robot, ghost, plant, object, or other character as a minimal square logo or app-icon artwork, including when the agent should infer three distinct IP directions from product-repository context before generating two-color or three-color logo candidates.
+description: Generate highly simplified personified IP mascot logos with Flat-first geometry, rounded heavy forms, two IP colors plus one background color by default, and extremely subtle neo-skeuomorphic shading. Use when creating an animal, creature, robot, ghost, plant, object, or other character as a minimal square logo or app-icon artwork, including when the agent should infer three distinct IP directions from product-repository context.
---
# IP as Logo
@@ -18,10 +18,10 @@ Create a logo first and a character second. Reduce the subject to a compact symb
4. Choose candidate subjects deliberately:
- When the user explicitly specifies an IP subject, keep that subject across the three candidates and vary only composition, silhouette treatment, secondary color region, or personality emphasis.
- When the user does not specify an IP subject, use three genuinely different IP subjects or metaphors. Tie each one to a different product attribute or brand promise; do not return three arbitrary animals with no rationale.
-5. Produce three candidates by default: candidate A is a two-color logo, while candidates B and C are three-color logos. If the user specifies a color count or number of candidates, follow it instead. Keep required product cues, identifying features, complexity limits, and any supplied palette consistent enough for useful comparison.
+5. Produce three candidates by default, each using three semantic colors in total: two IP colors plus one background color. Do not reserve any fraction of the default set for two-color logos. If the user explicitly specifies another color count or number of candidates, follow it instead. Keep required product cues, identifying features, complexity limits, and any supplied palette consistent enough for useful comparison.
6. Determine the available image-generation path before promising output. In Codex, use ImageGen when it is available. In any other agent environment, use an available configured image generator; if none is available, ask the user whether they can provide or enable one. Do not fabricate generated results.
7. If the runtime supports subagents, optionally assign one candidate to each subagent and run them in parallel. Give every subagent the same product brief, shared constraints, and one distinct IP or variation assignment. If subagents are unavailable, generate the candidates through separate image-generation calls or jobs.
-8. If the user supplies a background palette, reserve every supplied color for backgrounds unless they explicitly say otherwise. In a two-color logo, allow the background color to reappear only as negative-space facial cutouts, not as a separate painted IP region.
+8. If the user supplies a background palette, reserve every supplied color for backgrounds unless they explicitly say otherwise. Keep the two IP base colors distinct from the background. When the user explicitly requests a two-color logo, allow the background color to reappear only as negative-space facial cutouts, not as a separate painted IP region.
9. Abstract each subject using the complexity budget below. Generate every candidate independently; do not use existing logos or sibling candidates as image references when testing prompt-only reproducibility.
10. Inspect every output against every rejection rule. Retry with one targeted correction when practical; never hide a failed constraint with silent post-processing.
11. Preserve each model's native square output. Label the default set A, B, and C, then report each IP direction and rationale, saved path, prompt/color mapping, dimensions, opacity, and any remaining deviations. Ask which candidate the user wants to refine.
@@ -55,10 +55,9 @@ When proposing directions before generation, describe each in one compact line:
## Color and canvas
-- Define color modes by the total semantic colors in the finished logo, including the background:
- - **Two-color logo:** exactly one IP color plus one background color. Form eyes and mouth as negative-space cutouts that reveal the background; do not add a separate facial-feature color.
- - **Three-color logo:** exactly two IP colors plus one background color. Reuse one IP color for facial marks and keep the second IP color in one large continuous region.
-- Do not count lighter or darker shading variants as additional semantic colors when they remain within the limits below. Use more than three semantic colors only when the user explicitly requests them.
+- Default to exactly three semantic colors in the finished logo: two IP base colors plus one background color. Reuse one IP color for facial marks and keep the second IP color in one large continuous region. Do not mix in a default quota of two-color candidates.
+- Treat the two IP base colors as color families, not mechanically uniform flat swatches. Closely related highlight and shade variants are allowed for the ultra-light neo-skeuomorphic effect when they stay within the numeric limits below; these near-neighbor variants do not count as additional semantic colors.
+- Use more than three semantic colors only when the user explicitly requests them. Use a two-color logo only when the user explicitly requests it: exactly one IP base color plus one background color, with eyes and mouth formed as negative-space cutouts that reveal the background.
- Prefer a warm off-white such as cream or parchment over pure white, and charcoal or deep navy over pure black. Use pure black or white when the user requests it or when it provides the clearest result.
- Prefer backgrounds with a clear hue and restrained saturation: terracotta, muted coral, dusty plum, sage or forest green, glaucous or denim blue, ochre, and similar softened colors. Avoid neon, electric, candy-bright, and primary-color intensity unless explicitly requested. Also avoid reducing chroma until the color reads gray, muddy, or lifeless.
- Evaluate color in OKLCH when numeric control is available; do not use HSL saturation as the primary quality test. Use these default target bands:
@@ -67,8 +66,8 @@ When proposing directions before generation, describe each in one compact line:
- cream or parchment background: `L 0.92–0.98`, `C 0.01–0.06`.
- Treat `C < 0.05` on a chromatic background as likely too gray and `C > 0.20` as likely too saturated. These are defaults, not overrides for a user-supplied color.
- Maintain at least `3:1` relative-luminance contrast between the dominant IP silhouette and the background, and at least `4.5:1` between small facial marks and the surface beneath them. If the requested palette misses these targets, preserve the requested background and adjust the IP colors first.
-- In a three-color logo, build the second IP color from a large continuous region such as a face mask, hat, shell, belly, or visor. Do not scatter it into small decorative patches.
-- Keep highlight and shade variants in the same hue family: shift OKLCH hue by no more than about `5°`, chroma by no more than `0.02`, and lightness by roughly `±0.03–0.05` around the base color. Keep total peak-to-peak lightness variation at or below `0.10`.
+- Build the second IP base color from a large continuous region such as a face mask, hat, shell, belly, or visor. Do not scatter it into small decorative patches.
+- Keep every highlight and shade variant close to the IP base color it models: shift OKLCH hue by no more than about `5°`, chroma by no more than `0.02`, and lightness by roughly `±0.03–0.05` around that base color. Keep total peak-to-peak lightness variation within each IP color family at or below `0.10`. Do not introduce an unrelated hue under the label of shading.
- Keep the background visually solid. Across unobstructed background areas, target no more than about `0.02` OKLCH lightness variation and `0.01` chroma variation; reject visible vignettes or directional gradients. Report model drift rather than silently flattening it in post-processing.
- Require a fully opaque, edge-to-edge background. Do not generate transparency, an extra white border, outer frame, card, container, rounded App-icon mask, or artificially rounded image corners.
- Generate a direct `1:1` square with square outer corners. Request approximately `1536 × 1536`; accept and preserve a native `1254 × 1254` result when that is the service output limit. Never resample merely to reach the requested number.
@@ -77,11 +76,11 @@ When proposing directions before generation, describe each in one compact line:
```text
Create one highly simplified IP mascot logo candidate, not a character illustration.
-Background: fully opaque edge-to-edge solid <background>; use this color only for the background, except where two-color mode reveals it through negative-space facial cutouts.
+Background: fully opaque edge-to-edge solid <background>; use this color only for the background.
Subject: <subject> reduced to one rounded continuous silhouette and one defining feature.
Complexity: 6–10 basic shapes, at most two internal color regions, only two eyes and one mouth, readable at 32 × 32.
-Color count: exactly <two/three> semantic colors in the complete logo, including the background. For two colors, use one IP color and background-colored negative-space facial cutouts. For three colors, use two IP colors and keep the second IP color in one continuous region.
-Color behavior: softened but clearly chromatic background; warm off-white and charcoal/deep navy are preferred neutrals; silhouette/background contrast >= 3:1 and facial-detail contrast >= 4.5:1; keep shading in the same hue family.
+Color count: exactly three semantic colors in the complete logo: two IP base colors plus one background color. Reuse one IP color for facial marks and keep the second IP color in one continuous region.
+Color behavior: softened but clearly chromatic background; warm off-white and charcoal/deep navy are preferred neutrals; silhouette/background contrast >= 3:1 and facial-detail contrast >= 4.5:1. Allow only closely related highlight and shade variants around each IP base color for the ultra-light modeling; do not count those variants as new semantic colors.
Composition: upright, emerging from the lower-left or lower-right, filling 75–85%; show both paired identifying features.
Style: Flat-first with one diffuse internal highlight and one broad internal shade, 8–12% tonal variation, visible but extremely subtle.
Forbid: illustration detail, repeated anatomy, thin lines, sharp points, colors beyond the selected total count, pure flatness, strong 3D, clay, plastic, toy rendering, texture, gloss, bevel, external shadow, text, border, transparency, App mask, or rounded canvas corners.
@@ -90,7 +89,7 @@ Forbid: illustration detail, repeated anatomy, thin lines, sharp points, colors
## Reject the output when
- It reads as an illustration rather than a symbol, exceeds the complexity budget, or fails at small size.
-- The palette exceeds the selected total color count; a two-color logo adds a separate facial-feature color; or a three-color logo scatters its second IP color into decorative patches.
+- The default palette does not contain exactly two IP base colors plus one background color; the second IP color is scattered into decorative patches; shading introduces an unrelated hue; or an explicitly requested two-color logo adds a separate facial-feature color.
- A background-only color appears as a painted IP region rather than a negative-space cutout, or the background is transparent.
- A default chromatic background is neon or candy-bright, or is desaturated until it reads gray or muddy; small facial marks lack clear contrast.
- Any contour is thin, sharp, spiky, or visually fragile.