monolink-skill

monolink-skill.zip is a packaged Codex skill for generating monochrome poetic collage photography prompts or images.

The current version focuses on cross-domain visual translation: the main image and companion image must share a clear visual relationship, but they should not simply stay in the same subject, material, or scene category.

Core Idea

The skill follows this priority:

1. Cross-domain translation first
2. Visual relationship second
3. Monochrome photobook style third

It is designed to avoid weak same-domain pairings such as:

• floral dress -> flower field
• swallows -> sky or clouds
• sunset road -> another road or road texture

Instead, it pushes pairings such as:

• floral dress -> moth wings, old map islands, mildew stains, star clusters
• swallows -> feather barbs, stitching, human silhouettes, music notation
• sunset road -> palm lines, blood vessels, river paths, film scratches

Hard Rules

• The two images must cross at least two domains, preferably three: subject type, material, scale, scene system, or behavior.
• The relationship must remain visually readable through contour, direction, rhythm, structure, scale contrast, extension, or replacement.
• The white gutter between panels must stay completely clean. No subject, line, smoke, crack, shadow, petal, stroke, or visual bridge may enter or cross the gutter safety zone.
• The outer white border must be exactly equal on all four sides. If composition needs adjustment, resize or crop only the internal image area; never compensate by widening, narrowing, or cropping one side of the outer border.
• Black-and-white grain, wide borders, dust, scratches, and quiet mood are only the outer style layer. They must not compensate for a weak relationship.

Package Structure

monolink-skill/
├── SKILL.md
├── agents/
│   └── openai.yaml
└── references/
    ├── composition-grammar.md
    ├── examples.md
    ├── prompt-template.md
    ├── relation-first-rules.md
    └── style-analysis.md

Install

Extract monolink-skill.zip into your Codex skills directory.

The zip contains the complete skill folder and is the artifact intended for installation or sharing.

Generated Output Naming

AI-generated images created for this project should be saved under output/.

Before naming, inspect each image and identify its actual theme. Use this filename format:

YYYYMMDD-theme-##.png

• YYYYMMDD: generation or batch date, such as 20260525
• theme: Chinese image theme, such as 思念, 选择, or 高兴
• ##: two-digit sequence number, such as 01, 02, 14

Example: 20260525-思念-07.png