photo-abstract-editorial

`photo-abstract-editorial` is a Codex Skill designed to construct structured, vertical editorial layouts from user-supplied photos. Rather than applying generic neural style transfer, generative redraws, or heavy filters, this framework enforces a strict dual-component composition: a pristine original photo section paired with a derived abstract memory panel and poetic typography. ### Technical Architecture & Execution Model Under the hood, the skill operates as a parametric prompt framework tailored for Codex environments. It analyzes an input photograph to extract three primary visual vectors: spatial hierarchy, compositional rhythm, and palette dynamics. The core constraint of the framework dictates that the abstract panel must be fully traceable to real visual elements inside the uploaded photograph. Instead of hallucinating arbitrary geometry, the panel builds abstract motifs—such as color blocks, soft organic masses, arc strokes, stacked bands, or fine line work—directly out of the spatial facts found in the original shot. The repository structure isolates skill execution metadata from raw language prompts: ```text photo-abstract-editorial/ ├── SKILL.md # Skill workflow and execution constraints ├── agents/openai.yaml # Codex agent configuration & metadata └── references/ ├── photo-abstract-editorial-prompt.zh-CN.md └── photo-abstract-editorial-prompt.en.md ``` ### Parametric Customization The system functions as a modular layout engine. You can adjust several structural parameters within the prompts to modify output behavior: * **Canvas & Component Ratios:** Modify vertical height splits between the original photo and the abstract panel, adjust canvas aspect ratios, and control spatial white space around the abstract motifs. * **Palette Control:** Set background tones (such as default ivory), control color saturation, and cap the number of dominant vs. accent colors pulled from the source photo. * **Form Factors:** Select or mix specific visual elements, including short bars, simplified architectural massing, fine lines, or organic shapes. * **Abstraction Range:** Shift the visual weight between pure relational abstraction and identifiable structural features (e.g., retaining key outline details for recognizable landmarks or objects). ### Setup and Integration Workflow #### Method 1: Registering as a Codex Skill Copy the core repository into your local Codex skills directory: ```bash cp -r photo-abstract-editorial ~/.codex/skills/ ``` Once installed, open a new Codex session, upload a photograph, and invoke the skill using the following prompt: > 使用 `photo-abstract-editorial` 将这张照片制作成摄影与抽象面板组合的编辑作品。 The engine processes the photo by keeping the source image intact at the top, generating the abstract panel below it, and appending a single original English title (with an optional subtitle). #### Method 2: Direct Prompt Execution If you are running generation pipelines outside Codex's skill architecture, open either reference file directly and use the text as a base prompt for your workflow: * Chinese Reference: `references/photo-abstract-editorial-prompt.zh-CN.md` * English Reference: `references/photo-abstract-editorial-prompt.en.md` ### Licensing The project is released under the **CC BY-NC-SA 4.0** license. It is strictly limited to personal, educational, and non-commercial usage. Commercial distribution or integration into paid services is explicitly prohibited by the author.