Here's a prompt set for Biscuit. I assumed a soft, painterly picture-book style since you didn't name one. Swap it in the Style layer if you want something else.
Short prompt
A fluffy orange tabby cat sitting on a lighthouse balcony at sunset, seen from behind and slightly to the side, looking out at the sea. Warm children's picture-book illustration, portrait orientation.
Detailed prompt
A fluffy orange tabby cat with a thick, soft coat and darker orange stripes sits calmly on the iron railing-side floor of a lighthouse balcony, seen from behind at a three-quarter angle, with one ear turned toward the viewer and its tail curled around its paws. It gazes out over the sea. Setting: a white lighthouse with a red-trimmed balcony rail, a calm ocean below, a few small waves, and distant gulls, at sunset. Composition: vertical full-page illustration, the cat in the lower third on the left, a large expanse of sky and sea above and to the right, the horizon placed about one third from the top. Light: warm golden sunset light from the front, a rim of orange glow along the cat's fur, long soft shadows on the balcony floor. Style: warm, hand-painted children's picture-book illustration, soft gouache and watercolor textures, gentle rounded shapes, clear and friendly. Palette: glowing oranges, peach, and soft pink sky fading to deep teal-blue sea, with cream and red accents on the lighthouse. Mood: peaceful, wistful, cozy.
Variations (one change each)
- Composition: a wider view showing the whole lighthouse tower rising up the page with Biscuit tiny on the balcony, for a sense of scale.
- Light: the sun just touching the horizon, with the lighthouse lamp beginning to glow and the first stars in a purple-blue upper sky.
- Style: flat vector-style illustration with simple shapes, a limited palette, and a subtle paper-grain texture.
If the result is off
- Cat doesn't look like Biscuit or looks too realistic: edit the Style layer ("stylized, rounded, cartoon proportions") and the cat's details (add "big round eyes" or "chubby cheeks" if you want them).
- Cat faces the camera: repeat "seen from behind, looking away from the viewer" at the start of the prompt.
- Scene too busy: remove the gulls and waves, and keep only the cat, the rail, the sea, and the sky.
- Paws or tail look odd: ask for the paws tucked under the body (a "loaf" pose), which is easier for generators to draw.
- Colors muddy: name the palette more strictly (for example "only orange, pink, and teal").
To confirm
- Aspect ratio: the syntax differs by tool. For a portrait page, 2:3 or 3:4 is typical, but match your book's trim size. Check your tool's documentation for how to set it.
- Consistency across pages: most tools won't keep Biscuit identical from picture to picture. If you need him in more illustrations, keep this cat description word for word, and look at your tool's reference-image or character-consistency features.
- Licensing: check the tool's terms for commercial use and print publication.
- Bleed and text: leave space at the edges for trimming. If text will sit on the page, the empty sky area is a good spot. Add the text in a layout program, not in the generator.
If you tell me which tool you pick, or Biscuit's details (eye color, white markings, a collar), I can tailor the prompt.
Real run in an isolated folder with only this skill installed. Only the Skill and Read tools were enabled, so no command was run and no file was written. The scenario is fictional. No image was generated: the model only wrote text prompts, so how well they work depends on the image tool you use. The prompts were not tested in any image generator.
What it does
Builds an image prompt in layers: subject, details, setting, composition, light, style, mood and color, and technical settings. It asks about the purpose (illustration, product, social post, wallpaper) and the aspect ratio, puts the most important words first, keeps one idea per prompt and avoids contradictions and empty praise like "stunning, masterpiece, 8k". It handles the tricky parts: text inside images, hands and counts, negative instructions, and describing a look instead of naming a living artist. A reference file has a fill-in template, worked examples and a table of what to edit when the first result is off.
Output
A short prompt, a detailed prompt ready to paste, two or three variations that change one thing each, what to change if the result is off, and what to confirm (aspect ratio, tool syntax, usage terms).
Good for
Writing prompts for illustrations, portraits, scenes and concept art, for any image tool.
Low risk: pure instructions with no scripts, no network access and no file writes. It writes text prompts and does not generate images; results differ between tools, models and even runs, and tool-specific syntax must be checked in the tool's documentation. It will not help imitate a real person's likeness to deceive, make sexual content involving real people or minors, or mislead with fake evidence, and it warns against copying a living artist's signature style or a brand's protected assets. Check the tool's commercial-use terms. Original skill by AIBars (MIT). Tried once on an invented story illustration; the prompts were not run in an image generator.