lcqysl/gems/spatial

Skill: Spatial Intelligence & Layout

Description

This skill should be triggered when the user's request involves multiple objects, complex scene arrangements, or specific physical relationships between elements.

• Trigger when: The request includes terms like "on the left/right," "above/below," "behind/in front of," "taller than," "partially hidden," "jumping over," "in a circle," or when multiple distinct characters/objects need to interact in a specific setting.
• Do NOT trigger when: The request is for a single, centered subject with no background or specific spatial constraints.

Instructions

When this skill is active, rewrite the user's prompt explicitly covering all 10 spatial dimensions to ensure high-order spatial reasoning and physical consistency:

1. Object Category (S1): List all specific objects and components.
2. Object Attribution (S2): Define colors, materials, and textures for each object.
3. Spatial Position (S3): Use absolute coordinates (e.g., top-left, center) or relative positions (e.g., to the left of X).
4. Spatial Orientation (S4): Specify facing directions (e.g., facing left, upside down, profile view).
5. Spatial Layout (S5): Describe group arrangements (e.g., in a circle, a straight line, a "V" shape).
6. Spatial Comparison (S6): Define relative sizes or quantities (e.g., twice as large as X, three times taller than Y).
7. Spatial Proximity (S7): Describe physical distances (e.g., touching, far from, 5 inches apart).
8. Spatial Occlusion (S8): Detail 3D layering and depth (e.g., X partially obscuring Y).
9. Spatial Motion (S9): Capture dynamic states (e.g., mid-air, jumping over, liquid splashing).
10. Spatial Causal Interaction (S10): Link causes to effects (e.g., wind blowing a cloak, a ball hitting water to create ripples).

Writing Principles:

• Logical Soundness: Ensure no cyclical layout errors (e.g., avoid "A is left of B, B is left of A").
• Explicitness: Replace vague words like "nearby" with "standing 2 feet away on the right side".
• Format: Output the result as a single, fluent, and descriptive paragraph. Do not use bullet points in the final prompt.

Output Format

Return ONLY the final enhanced prompt text. Do not include any conversational filler, introductory remarks, or prefixes like "Enhanced prompt:".