Codex Skill Extracts Abstract Relationships While Preserving Original Photos

Photo Abstract Editorial is a Codex skill that creates abstract panels by extracting spatial relationships, composition, and color roles from a photo, rather than directly modifying its pixels.

Unlike existing filters or style transfers, this skill preserves the original photo as-is, producing vertical editorial works that combine abstract memory panels derived from the photo with English titles.

Key Features and How It Works:

• Relationship-Centric Abstraction: Instead of simply reducing forms, it extracts only visual 'relationships' such as spacing between subjects, layers of light, and structural axes, then reconstructs them as marks.
• Strict Guardrails: All abstract marks must be grounded in the visual facts of the original photo, strictly prohibiting the addition of decorative elements or artificial patterns.
• Systematic Process: Processes information through the stages of Deconstruct $\rightarrow$ Selective Preservation $\rightarrow$ Abstract $\rightarrow$ Reconstruct.
• Structured Prompts: Controls the workflow and consistency of outputs through rule documents (SKILL.md) and sophisticated prompts instead of code.

Usage and License:

• Copy it into the Codex skill directory to use, or use the provided prompts directly with image generation tools.
• The license is limited to personal, educational, research, and non-commercial use; prior permission from the author is required for commercial use.