Cover 3d Eye

3D Eye Cover

Use this skill to create a finished image prompt, or call an image generation tool directly when the user asks to generate the final cover. The result should feel like a premium dark-mode tutorial poster for running local AI: bold, private, terminal-native, slightly rebellious, and easy to understand at social-feed speed.

This is the visual style member of the 3d-eye family. Use it directly for covers/posters. When the same tutorial also needs step-by-step diagrams, hardware maps, privacy comparisons, or quantization explainers, route planning through article-visual-planner and pass this skill as the selected style, for example /article-visual-planner:cover-3d-eye.

Output Type

Use the explicit --out-type parameter to decide what to output.

• --out-type template: output the invocation template only.
• --out-type prompt: output the final image prompt only.
• --out-type all: output the template first, then the final image prompt.
• Omitted --out-type: default to template.

Treat 直接生成, 生成海报, 生成封面, 出图, and 生成图片 as direct image generation when an image tool is available.

Template Output

For template-only mode, output exactly this structure and fill every field:

使用 $cover-3d-eye 生成一张封面
主题词：{topic}
副标题：{subtitle}
画幅比例：{ratio}
语言：{language}
用途：{use_case}
核心钩子：{hook}
视觉主体：{hero_subject}
补充背景：{context}
禁用元素：{forbidden_elements}

Required Inputs

Extract or infer:

• Topic / main title: required.
• Subtitle: optional.
• Aspect ratio: optional; infer from use case when omitted.
• Language: Chinese, English, or mixed Chinese-English.
• Use case: X cover, WeChat cover, blog header, tutorial cover, YouTube thumbnail, product education poster, knowledge card.
• Core hook: privacy, uncensored, offline, local ownership, hardware fit, quantization, speed, cost, safety mode off, no cloud, no account, no logs.
• Hero subject: terminal window, lock, hardware card, local-vs-cloud split, workflow strip, 3D Eye mascot, torn censorship tape, model file, API endpoint.
• Extra context: optional.
• Forbidden elements: optional; combine with style defaults.

Aspect Ratio Defaults

• X article cover / X header: 5:2.
• WeChat article cover: 2.35:1.
• Blog header / tutorial cover: 16:9.
• YouTube thumbnail: 16:9.
• Knowledge card / social post: 1:1.
• Poster: 3:4 or 4:5.

State the exact ratio and canvas size in the first line of prompt-only output.

Visual System

Use these rules strictly:

• Background: deep black or near-black #050707, subtle blueprint grid, faint scan lines, corner registration marks, restrained film grain.
• Color: black, off-white title text, neon terminal green #39FF73, dim gray secondary text, small red warning accents only when the subject needs error/risk.
• Green usage: terminal command text, active borders, arrows, checkmarks, selected cards, glow around the lock/mascot/focus item.
• Typography: huge condensed bold sans title, cream-white, poster-like; secondary text can use monospaced terminal style.
• UI language: rounded terminal windows, command prompts, CLI cursors, step cards, hardware memory cards, lock/shield icons, cloud/local comparison panels, API endpoint labels.
• Mascot system: optional but recommended for series continuity. Use the 3D Eye mascot as a friendly hand-drawn white eyeball character with a rich blue iris, black sketch outlines, thick expressive eyebrows, black rubber arms/legs, white cartoon gloves, and small black shoes. It should feel intelligent, warm, focused, and action-oriented rather than robotic. Preferred poses: waving welcome, running forward, pointing direction, holding a blue skill card, thumbs-up, open palm, or confident center pose. Keep the mascot blue-white-black with small light-blue motion marks; do not add headset/cap/tool-belt variants unless the user explicitly asks.
• Composition: strong title layer, one clear hero scene, 1-3 supporting UI/diagram elements, generous dark negative space, bright green focus path.
• Mood: private, local, hacker-clean, educational, confident, a little defiant.

Avoid generic cyberpunk city backgrounds, purple/blue neon, RGB gaming rooms, Matrix code rain, robot faces, stock hacker hoodies, photoreal people, cluttered dashboards, random circuit-board backgrounds, cheap 3D icons, illegible microtext, official product logos unless the user explicitly asks, and copying existing case-image text verbatim.

Cover Composition Modes

Pick one mode unless the user specifies one:

• Censored breakthrough: huge title, torn black tape, green crack/glow, lock or terminal revealing the thesis.
• Local ownership hero: terminal plus unlocked lock/shield, with install-run-own workflow strip.
• Split world: local device vs cloud server, with local highlighted and cloud dimmed.
• Mascot explainer: 3D Eye mascot points to a terminal, card, or lock while title carries the argument.
• Hardware promise: memory/VRAM card, model size, speed tag, local machine emphasis.
• Danger/mistakes poster: red warning bullets with mascot reaction, but keep green as the dominant accent.

Prompt Output

For prompt-only mode, produce a provider-neutral image-generation prompt. Do not mention provider names, model names, or runtime-specific syntax unless the user explicitly asks.

Include:

1. Exact canvas ratio and size.
2. Topic, subtitle, use case, language, and core hook.
3. Selected composition mode and why it fits.
4. Hero subject and 1-3 concrete visual anchors.
5. Title hierarchy and text placement plan.
6. Background, terminal UI, neon green color system, typography, mascot usage.
7. A concise avoid list.

If generating directly, keep all analysis internal and output only the generated image result.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: imartinstudio
• Source: imartinstudio/cover-prompt-skills
• License: MIT

Install and usage instructions live in the source repository linked above.