Validate images against tacosdedatos illustration style guide. Use this skill when reviewing generated images, checking uploaded images for brand consistency, providing quality feedback on visual assets, or deciding if an image is ready for publication. Returns pass/fail with specific actionable feedback.
This skill inherits all available tools. When active, it can use any tool Claude has access to.
Quality gate for tacosdedatos illustrations.
Reference: agents/shared/illustration-style-guide.md for authoritative mode definitions and color palette.
First, determine which of the 7 modes the image fits (or should fit):
| Mode | Key Identifiers |
|---|---|
| Abstract/Geometric | Shapes, lines, no figures, technical feel |
| Paper-Cut/Layered | Overlapping silhouettes, depth layers |
| Surrealist/Evocative | Dreamlike, unexpected combinations, emotional |
| Nature + Tech Fusion | Plants + data cards, organic + digital |
| Atmospheric/Cinematic | Scene-based, moody lighting, environment |
| Playful Cartoon | Expressive characters, dynamic, icons |
| Data Dashboard | Charts, visualizations, laptop/screens |
If no mode fits: Flag as potentially off-brand.
Check each criterion:
| Criterion | Check | Pass/Fail |
|---|---|---|
| Mode fit | Does it clearly fit one of the 7 modes? | |
| Color palette | Are colors from the tacosdedatos palette? | |
| Texture | Is there subtle grain/risograph texture? | |
| No text | Is there any embedded text, labels, or watermarks? | |
| Metaphor clarity | Is there a clear visual concept? | |
| Negative space | Is there generous breathing room? | |
| Aspect ratio | Is it landscape (~16:9 or 3:2)? |
Check against palette:
| Color | Hex | Present? |
|---|---|---|
| Negro Mezquite (bg) | #121212 | |
| Teal Oscuro (bg) | #1a2e35 | |
| Crema Nevada | #F1FAEE | |
| Rojo Fuego | #E63946 | |
| Azul Niebla | #A8DADC | |
| Naranja Cempasúchil | #F4A261 | |
| Verde Maguey | #4E937A |
Tolerance: Colors don't need to be exact hex matches, but should be recognizably from the palette family. Flag if:
| Style | Acceptable? |
|---|---|
| Silhouette (solid, no features) | ✓ |
| Layered profile (overlapping heads) | ✓ |
| Flat character (simple shapes) | ✓ |
| Expressive cartoon (big eyes, gestures) | ✓ |
| Abstract avatar (icon-style) | ✓ |
| Atmospheric figure (stylized in scene) | ✓ |
| Photorealistic human | ✗ |
| Stock illustration style | ✗ |
| Generic corporate people | ✗ |
Scoring:
Verdicts:
| Score | Verdict | Action |
|---|---|---|
| 7/7 | PASS | Ready for publication |
| 5-6/7 | PASS WITH NOTES | Minor issues, acceptable |
| 3-4/7 | NEEDS REVISION | Regenerate with fixes |
| 0-2/7 | FAIL | Start over with different approach |
Provide structured feedback:
## Image Validation Report
### Verdict: [PASS / PASS WITH NOTES / NEEDS REVISION / FAIL]
**Score**: X/7
### Mode Identified
[Mode name] - [Confidence: High/Medium/Low]
### Checklist Results
| Criterion | Result | Notes |
|-----------|--------|-------|
| Mode fit | ✓/✗ | |
| Color palette | ✓/✗ | |
| Texture | ✓/✗ | |
| No text | ✓/✗ | |
| Metaphor clarity | ✓/✗ | |
| Negative space | ✓/✗ | |
| Aspect ratio | ✓/✗ | |
### Issues Found
1. [Specific issue]
2. [Specific issue]
### Recommended Fixes
1. [Actionable fix]
2. [Actionable fix]
### Prompt Adjustment (if regenerating)
[Specific prompt changes to make]
Detection: Any letters, numbers, words, labels visible
Fix: Add to negative prompt:
Absolutely no text, no letters, no words, no writing, no labels, no numbers, no watermarks
Detection: Colors outside palette (bright, pastel, neon)
Fix: Add explicit hex codes:
Colors MUST use only: #121212, #F1FAEE, #E63946, #A8DADC, #F4A261, #4E937A
Detection: No clear focal point, busy composition
Fix: Simplify prompt:
Single focal element. Generous negative space. Minimal composition. Clean and simple.
Detection: Flat, plastic, or glossy appearance
Fix: Add texture cues:
Subtle risograph grain texture throughout. Matte finish. No glossy surfaces.
Detection: Looks like a photo or 3D render
Fix: Emphasize style:
Flat digital illustration. No photorealism. No 3D render. No realistic textures.
Detection: Looks like corporate clip art or stock illustration
Fix: Be more specific about style:
In the style of modern editorial illustration. Sophisticated and unique. Not corporate clip art.
Detection: Square or portrait orientation
Fix: Specify in generation settings:
aspect_ratio="16:9" or aspect_ratio="3:2"
For fast validation without full report:
## Quick Check: [Title]
**Verdict**: [PASS/FAIL]
**Key Issues** (if any):
- [Issue 1]
- [Issue 2]
**Fix**: [One-line recommendation]
Image: Abstract geometric shapes flowing into brain network
## Image Validation Report
### Verdict: PASS
**Score**: 7/7
### Mode Identified
Abstract/Geometric - Confidence: High
### Checklist Results
| Criterion | Result | Notes |
|-----------|--------|-------|
| Mode fit | ✓ | Clear geometric/abstract style |
| Color palette | ✓ | Uses #121212, #F1FAEE, #E63946, #A8DADC |
| Texture | ✓ | Subtle grain visible |
| No text | ✓ | Clean, no text |
| Metaphor clarity | ✓ | Clear AI/processing metaphor |
| Negative space | ✓ | Good breathing room |
| Aspect ratio | ✓ | Landscape 16:9 |
### Issues Found
None
### Recommended Fixes
None - ready for publication
Image: Cartoon character but with text label
## Image Validation Report
### Verdict: NEEDS REVISION
**Score**: 5/7
### Mode Identified
Playful Cartoon - Confidence: High
### Checklist Results
| Criterion | Result | Notes |
|-----------|--------|-------|
| Mode fit | ✓ | Matches playful cartoon style |
| Color palette | ✓ | On-brand colors |
| Texture | ✗ | Too smooth, needs grain |
| No text | ✗ | Text label visible on character |
| Metaphor clarity | ✓ | Clear concept |
| Negative space | ✓ | Good composition |
| Aspect ratio | ✓ | Correct landscape |
### Issues Found
1. Text label "AI" visible on character's shirt
2. Missing subtle texture/grain
### Recommended Fixes
1. Regenerate with stronger "no text" negative
2. Add "subtle risograph grain texture" to prompt
### Prompt Adjustment
Add: "Absolutely no text, no letters, no writing. Subtle risograph grain texture throughout."
This skill can be called:
The validator does not generate images—it only evaluates them.