360 Secure SkillHub
360 安全技能中心

Aura Image Gen

提供 8 种视觉风格、8 种布局与 5 种平台比例的图片生成指引,含星宿、塔罗等组合预设和 prompt 拼装流程,输出生成方案与生图 prompt 文本;实际出图需用户自备 AI 生图工具。

luoboask
luoboask
热度 48.48
最近更新 2026-05-11 17:36:26
分类:创作依赖:内部调用大模型应用场景:图像生成 / 社交媒体 / 模板

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add aura-image-gen
Metadata
name
aura-image-gen
version
2.0.0
description
Multi-platform infographic card generator for Xiaohongshu, Pinterest, Instagram, and more. Provides 8 visual styles × 8 layouts with platform-specific aspect ratios and prompt templates. Use when user mentions '生成图片', 'AI 画图', 'generate image', 'draw', '小红书卡片', 'Pinterest pin', 'Instagram post', or needs prompt guidance for creating social media visuals with any AI image tool.

Aura Image Generator (多平台图片生成)

Generate eye-catching social media card series. Style × Layout × Platform three-dimensional system.

Three Dimensions

Dimension Controls Options
Style Visual aesthetics mystic, cute, notion, bold, retro, ink-wash, tarot, minimal
Layout Information structure sparse, balanced, dense, list, comparison, flow, card, spread
Platform Aspect ratio + safe zones xiaohongshu, pinterest, instagram-square, instagram-portrait, youtube-thumbnail

Platform Specs

Platform Ratio Pixels Safe Zone Notes
xiaohongshu (默认) 3:4 1080×1440 避开右上角互动区、底部标题栏
pinterest 2:3 1000×1500 标题前 40 字符最重要,留足顶部空间
instagram-square 1:1 1080×1080 Feed 标准,中心聚焦
instagram-portrait 4:5 1080×1350 Feed 竖版,最大高度
instagram-story 9:16 1080×1920 Story/Reels,避开上下边缘
youtube-thumbnail 16:9 1280×720 大字 + 高对比,小图可读

Style Gallery

Style Description Best For
mystic (Default for astrology) 深色星空 + 金色点缀,神秘玄学感 星宿合盘、塔罗解读
cute 甜系粉嫩,经典小红书风 星座日常、轻松科普
notion 极简线条手绘,知识感 关系对照表、知识卡片
bold 高对比强冲击,大字报 避坑指南、排行榜
retro 复古怀旧,做旧质感 传统文化、国风内容
ink-wash 水墨国风,中式美学 星宿神兽、传统占卜
tarot 塔罗牌艺术风,新艺术运动边框 塔罗解读、牌面赏析
minimal 极简高端,大量留白 金句、单牌解读

Style: Mystic (详细定义)

canvas:
  ratio: portrait-3-4
  grid: single | dual

colors:
  primary: ["#0A0E2A", "#1A1040"]  # 深夜蓝/深紫
  accent: ["#C9A96E", "#D4AF37"]   # 古铜金/正金
  text: ["#E8D5B7", "#FFFFFF"]     # 暖白/纯白
  glow: ["#6B5CE7", "#A78BFA"]     # 紫色光晕

elements:
  background: starfield | nebula-gradient
  decorations: [constellation-lines, star-sparkles, moon-phases]
  frames: gold-border | arch-frame | none
  emphasis: glow-highlight | gold-underline

typography:
  title: serif-elegant | brush-calligraphy
  body: clean-sans | handwritten
  numbers: display-large

Style: Tarot (详细定义)

canvas:
  ratio: portrait-3-4
  grid: single

colors:
  primary: ["#1B0A2E", "#2D1B4E"]  # 深紫
  accent: ["#D4AF37", "#B8860B"]    # 金
  highlight: ["#8B0000", "#4A0080"] # 深红/紫
  text: ["#F5E6CC", "#FFFFFF"]

elements:
  background: velvet-dark | ornate-pattern
  decorations: [art-nouveau-border, celestial-symbols, card-frame]
  frames: tarot-border | ornate-gold
  emphasis: gold-foil | emboss

typography:
  title: art-nouveau | display-serif
  body: elegant-serif
  card-name: all-caps-spaced

Style: Ink-wash (详细定义)

canvas:
  ratio: portrait-3-4

colors:
  primary: ["#1A1A1A", "#4A4A4A"]  # 墨色
  accent: ["#8B0000", "#C41E3A"]    # 朱红
  background: ["#F5F0E8", "#E8DCC8"] # 宣纸色
  text: ["#2C2C2C", "#8B0000"]

elements:
  background: rice-paper-texture | ink-splash
  decorations: [brush-strokes, seal-stamp, ink-dots]
  frames: scroll-border | none
  emphasis: red-seal | brush-circle

typography:
  title: brush-calligraphy
  body: kai-style

Layout Gallery

Layout Density Best For
sparse 1-2 points 封面、金句、单牌
balanced 3-4 points 标准内容、性格解读
dense 5-8 points 知识卡片、对照表
list 4-7 items 排行榜、清单
comparison 2 sides 两人合盘、正逆位对比
flow 3-6 steps 流程、时间线
card Single focus 单张塔罗牌/星宿卡
spread Multi-position 牌阵展示

Language Support

Language When Notes
中文 小红书内容 默认,所有文字元素用中文
English Pinterest / international 所有文字元素用英文
双语 用户指定 同一张图可以中英双语(标题中文 + 副标题英文)

Prompt language rule:

  • AI 生图的 prompt 始终用英文(模型理解最好)
  • 图片内的展示文字跟随目标平台语言
  • Prompt 中用 Text on image in Chinese: "xxx" 或 Text on image in English: "xxx" 明确指定

Presets (风格 + 布局 + 平台组合)

星宿内容预设

Preset Style + Layout + Platform Use Case
xingxiu-profile mystic + balanced + xiaohongshu 单宿性格解读
xingxiu-compatibility mystic + comparison + xiaohongshu 两人合盘
xingxiu-ranking bold + list + xiaohongshu 关系排行榜
xingxiu-lookup notion + dense + xiaohongshu 查宿对照表
xingxiu-beast ink-wash + card + xiaohongshu 星宿神兽卡

塔罗内容预设

Preset Style + Layout + Platform Use Case
tarot-single tarot + card + xiaohongshu 单牌解读
tarot-spread tarot + spread + xiaohongshu 牌阵解读
tarot-daily mystic + sparse + xiaohongshu 每日一牌
tarot-guide notion + list + xiaohongshu 新手教程
tarot-compare tarot + comparison + xiaohongshu 正位 vs 逆位

Pinterest 预设

Preset Style + Layout + Platform Use Case
pinterest-astrology mystic + balanced + pinterest 星座科普
pinterest-tarot tarot + card + pinterest 塔罗解读
pinterest-quote minimal + sparse + pinterest 金句灵感
pinterest-tutorial notion + flow + pinterest 教程步骤

通用预设

Preset Style + Layout + Platform Use Case
knowledge-card notion + dense + xiaohongshu 干货知识卡
warning bold + list + xiaohongshu 避坑指南
quote minimal + sparse + xiaohongshu 金句封面

Workflow

Step 1: Analyze Content

  • 识别内容类型(星宿/塔罗/知识/排行)
  • 识别目标平台(默认 xiaohongshu)
  • 推荐 preset 或 style + layout + platform 组合
  • 确定图片数量(封面 + 内容 + 结尾)

Step 2: Confirm

  • 展示推荐方案
  • 用户确认或调整

Step 3: Generate Outline

每张图的:

  • 位置(封面/内容/结尾)
  • 文字内容
  • 视觉概念
  • 使用的 layout + platform ratio

Step 4: Generate Images

  • 为每张图组装 prompt(style + layout + platform + content)
  • 参考 references/workflows/prompt-assembly.md 拼装完整 prompt
  • 使用任意 AI 生图工具执行(Midjourney/SD/DALL-E/通义万象等)
  • 第一张作为视觉锚点,后续用第一张做参考保持风格一致
  • 每张生成后报告进度

Step 5: Output

Series Complete!
Topic: [topic]
Platform: [platform] · Ratio: [ratio]
Style: [style] · Layout: [layout]
Images: N total
Files: 01-cover.png, 02-content.png, ...

Prompt Assembly

每张图的 prompt 结构:

Create a social media infographic card:

## Specs
- Platform: {platform}
- Aspect Ratio: {ratio} ({pixels})
- Style: {from style definition}

## Visual Rules
{from style yaml: colors, elements, typography}

## Layout
{from layout definition: density, structure, whitespace}

## Platform Safe Zones
{from platform specs: avoid UI overlays, title bars}

## Content
Position: {cover/content/ending}
Text: {actual content}
Visual Concept: {description}

## Language
{Match content language}

Visual Authenticity Tips

  • Add slight asymmetry to compositions
  • Offset text 2-3px from perfect center
  • Include subtle texture noise in backgrounds
  • Vary line thickness naturally
  • Allow soft edges on color blocks

References

Presets (style definitions — load when generating):

  • references/presets/mystic.md — 玄学神秘风
  • references/presets/tarot.md — 塔罗艺术风
  • references/presets/ink-wash.md — 水墨国风
  • references/presets/cute.md — 甜系少女风
  • references/presets/notion.md — 极简知识风
  • references/presets/bold.md — 高冲击力风
  • references/presets/retro.md — 复古怀旧风
  • references/presets/minimal.md — 极简高端风

Elements (canvas specs):

  • references/elements/canvas.md — 画布尺寸/安全区/布局网格
  • references/elements/platforms.md — 多平台规格详解

Workflows (process guides):

  • references/workflows/prompt-assembly.md — Prompt 拼装指南
  • references/workflows/multi-platform.md — 多平台适配指南