Compile a theme, sentence, object, emotion, or article summary intoa single"Quiet Paper Specimen" cover prompt that can be handed directly to an image model.
Style anchor:one specimen laid out alone on aged paper—large areas of negative space, a single imaginable subject, restrained typography, one highly saturated color dot, print/scan paper texture.
Core principle:The style is locked down, the solution is left open; by default output a finished, decision-made prompt; when the user wants a template, output a copy-and-use structure with editable Chinese fields first and the fixed English style after.
Negative space is the protagonist, not the background. The entire tension of this skill comes from the relationship between "emptiness" and "one quiet dot." Any urge to fill the frame, add multiple subjects, add multiple colors, or add long passages of text must be held back.
Read
Read by task:
- Generate prompt: read
references/compile-contract.mdandreferences/module-library.md。 - User wants examples or theme inspiration: read
references/examples.md(if it exists).
Input
Infer from user input first; do not require a form to be filled out:
- Input mode: text theme, one sentence, one object, one emotion, an article summary, or a style assessment of an existing image.
- Core image: must land ononeimaginable subject or small visual cluster, chosen from the core image library in
module-library.md; the subject occupies 12-25% of the frame and is legible at thumbnail size. Abstract concepts must first be translated into concrete objects from the library (a chipped enamel cup, a straightened paper clip, a cicada shell—not "growth" or "connection").Do not repeatedly use leaves, old photos, coins, or paper boats—those are only one corner of the image library. - Text: main title defaults tolarge handwritten Chinese characters(brush/pen/pencil handwriting, character height about 10-18% of the frame height, readable at thumbnail size); optionally add one line of archival-style microtext or a small subtitle. When the user specifies Chinese, the original Chinese must be kept and must not be changed to English by default; secondary text stays quiet and sparse, but the main title must be clear—a title that blurs into the paper counts as a defect.
- Aspect ratio: default 3:5 vertical (paper pages are naturally vertical). When the user specifies another ratio, follow it first; landscape and square are also allowed, but still keep the large-negative-space logic.
Workflow
- Identify the input: determine whether it is text generation, a single object, emotion-oriented, an article cover, or a style assessment.
- Extract one core image: draw onlyonevisualizable idea from the user's content, and pick the corresponding object from the core image library. Better quiet than cramming in multiple elements; but quiet does not mean invisible.
- Choose modules: from
module-library.mdChoose the layout, anchor, headline, typography, texture, color accent, and mood. When there are multiple images, the core imagery of each must come fromdifferent families, and at least three other types of modules must also be varied. - Compile the prompt: output according to
compile-contract.md. By default, output the decided English art direction, leaving no{}, slash menus, either/or layouts, or undecided options. When the user explicitly requests a "copy-and-use template / reusable template / article prompt structure," output a template with the Chinese input fields first and the English style anchoring after. - Output: the selected modules + one complete prompt + one sentence explaining the choice. When the user only wants the prompt, the explanation may be omitted, keeping only the module summary and the complete prompt.
Routing Rules
- One sentence / one word: prioritize a minimalist specimen-style layout, with a single subject and sufficient white space, and keep only one line of edge-aligned micro-text for typography; the subject must still be clearly visible.
- Mood / memory / essay-oriented: prioritize soft anchors such as faded photos, semi-transparent geometry, and texture windows; color accents should lean toward low-saturation warm tones, and textures toward soft photocopy or scanned aged paper.
- Article cover: by default, provide a large Chinese handwritten headline; you may add one line of a small subtitle (archive micro-text or a short edge-aligned sentence), and the subtitle must be noticeably smaller than the main headline, with no information-dense cover.
- Specific object: preserve the object's silhouette and one or two most recognizable details, simplify it into a quiet hand-drawn/printed specimen, and do not create the impact of a product advertisement.
- Uploaded image redraw: preserve the original subject's identity and silhouette, reduce it to a single specimen on aged paper, and remove the original image's scene, lighting effects, and background clutter.
Output Format
Default finished prompt:
## 已选模块
- 输入模式：
- 核心意象（家族）：
- 布局模块：
- 锚点模块：
- 标题模块：
- 排版模块：
- 纹理模块：
- 色点模块：
- 情绪模块：
## 完整提示词
```text
...
```
## 选型理由
...
Copy-and-use template:
```text
文章主题或摘要：
{输入文章主题、标题、摘要，或直接粘贴文章内容}
用途和比例：
{例如：文章封面，竖版 3:5}
指定主标题或微文本：
{可选；没有就写：自动判断}
指定副标题文字：
{可选；没有就写：自动判断或不加}
特殊要求：
{指定核心意象、色点颜色、不能出现的元素；没有就写：无}
Create a quiet paper-specimen editorial cover from the theme above.
Style lock:
...
```
Do not add explanatory lines like "you only need to change this" or "the following content does not need to be changed" to the template. The fields themselves are already clear enough.
Quality Gates
- It must be a quiet paper specimen / independent magazine editorial logic, not a full-bleed poster, commercial advertisement, or infographic.
- The aged-paper white space must be the protagonist of the image: negative space occupies 70%-80%, and the subject occupies 12%-25%.
- Quiet does not mean invisible: the subject, color accent, and headline must all be recognizable at thumbnail size. A subject or headline blurred into the paper counts as a defect, not as sophistication.
- There must be onlyoneimaginable core image or small visual cluster, with no piling up of multiple subjects; the imagery must be selected from the core imagery library, and should not always fall back to leaves/old photos/coins/paper boats.
- There must be onlyonehigh-saturation color accent, occupying about 1.5%-2.5% of the canvas, and it must fall on the subject rather than float on the empty paper; the rest of the image is paper color, ink color, and low-saturation gray tones.
- The main headline by default uses large Chinese handwritten characters: the user's original text, character height about 10%-18% of the image height, with brushstrokes having a genuine handwritten irregularity; do not use computer handwriting font libraries, and do not make a bold black printed headline dominate the screen.
- Secondary text stays sparse and quiet: fragmented letters, edge-aligned micro-text, archive labels, and ghost text at most one or two places. Ghosting and fragmentation are only used for secondary text, never on the main headline.
- When the user requests Chinese, the user-specified original text must be used; avoid fake Chinese, garbled characters, and random labels.
- The subtitle may only be one line, noticeably smaller than the main headline; for Chinese, 6-12 characters are recommended, and for English, 2-6 words.
- The texture must be like real printed/scanned material: photocopying, halftone dots, silkscreen bleed, film grain, letterpress impression, mottled aged paper, pencil gray; not a post-production dirty filter or digital distressing.
- When output as a group, the core imagery of each piece must come fromdifferent families, and at least three other modules must be clearly different as well. Five pieces that are all paper scraps and plant specimens, even with different layouts, still count as a failure.
- Prohibited: full-bleed scenes, commercial poster impact, 3D/neon/cartoon/gradients, multiple objects, multiple colors, large blocks of body text, clean vector feel of modern UI, studio lighting, photorealistic product rendering.
- Do not fabricate already-generated image results; this skill only produces prompts or style evaluations.