원클릭으로 Manus에서 모든 스킬 실행

encyclopedia-card

스타4

포크0

업데이트2026년 5월 11일 09:50

Generate premium vertical encyclopedia-style knowledge cards / reference handbook pages / educational infographic cards for a single topic — like a publishable natural-history guidebook page mixed with modern editorial infographic design. Triggers on phrases like 百科卡 / 百科全书卡 / 博物卡 / 图鉴卡 / 知识卡 / 参考手册 / 手账卡片 / encyclopedia card, knowledge handbook card, field guide page, educational infographic card, reference card series. Signature look: one hero subject illustration + 2-4 zoomed-in detail callouts + 4-7 rounded modular info panels + quick rating / top-5 / summary chips, soft cream background, gentle shadows, small refined line icons, high density but calm reading rhythm. Do NOT use for 爆款新闻封面 (use tech-news-cover), 电影海报 (use editorial-poster), 小红书系列 (use xhs-card-series), dense business consulting report (use knowledge-infographic consulting-report preset), 单纯文章封面 (use article-cover-image), 漫画 (use comic-strip).

설치

Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.

Manus에서 실행

출처

inspirepan

inspirepan/nano-banana-playground

GitHub 저장소 열기 Creator 저장소 보기

다운로드

Manus에서 실행

관련 직업SOC

SOC 직업 분류 기준

그래픽 디자이너예술, 디자인, 엔터테인먼트, 스포츠, 미디어직·SOC 27-1024

SKILL.md

readonly

이 저장소의 다른 Skills

같은 저장소

article-cover-image

inspirepan/nano-banana-playground

Generate article cover images / blog hero images / 公众号封面 / 文章配图 with a 5-dimension system (type, palette, rendering, text, mood) plus font selection. Use this skill whenever the user asks to "generate cover image", "create article cover", "make a blog hero", "design a banner for this post", or in Chinese "生成封面 / 文章封面 / 公众号封面 / 配图 / 头图 / 题图". Supports cinematic 2.35:1, widescreen 16:9, square 1:1, and 4:3 aspect ratios. Do NOT use this skill for: 小红书多图卡片 (use xhs-card-series), 漫画分镜 (comic-strip), 知识科普长信息图 (knowledge-infographic), 单张电影场景 (scene-cinematic), 海报/印刷品 (editorial-poster); also skip for product photo, avatar, logo design, icon set, or pure text rewriting.

2026-05-144

comic-strip

inspirepan/nano-banana-playground

Multi-panel comic strip and storyboard generator. Use when the user wants to turn a story, joke, dialogue, tutorial bit, or short scene into a comic strip, manga panels, story panels, graphic novel page, storyboard, 漫画, 分镜, 四格漫画, 故事板, 短漫. Supports two output modes — single-sheet (one image with 2/3/4/6 panels) or multi-image (N sequential images linked by an anchor reference). Do not use for single cover art, infographics, posters, or single illustrations.

2026-05-144

editorial-poster

inspirepan/nano-banana-playground

Generate screen-print and duotone style editorial posters for movie posters, book cover posters, concert posters, cultural commentary covers, and limited-edition art prints. Use whenever the user mentions movie poster, book cover poster, editorial poster, screen-print, silkscreen, duotone, halftone, concert poster, alternative poster art, Mondo style, or 影评海报、书评封面、电影海报、演唱会海报、文化评论封面、丝网印刷海报、双色调海报. Strong visual core is screen-print silhouettes plus duotone color blocks plus halftone grain plus typography-as-composition. Do NOT use for slide decks, flowcharts, infographics, product photography, UI screenshots, or photorealistic key art.

2026-05-144

flat-illustration

inspirepan/nano-banana-playground

Generate layered 2.5D flat design illustrations with clean shapes, visible depth, cel-shaded planes, soft top-right lighting, balanced spacing, plain backgrounds, and restrained palettes. Use when the user asks for 扁平插画, flat illustration, layered vector illustration, Live2D-like flat character art, simple geometric character/object art, or cute clean icon-like scenes. Do NOT use for UI mockups, product ads, comics, dense infographics, cinematic scenes, realistic photography, logos, or text-heavy posters.

2026-05-144

knowledge-infographic

inspirepan/nano-banana-playground

Generate a single dense knowledge card / infographic / concept map / cheat sheet / business analysis report from a topic, outline or article. Trigger on "knowledge card", "infographic", "concept map", "summary card", "cheat sheet", "mindmap card", "quadrant card", "business report infographic", "干货卡", "知识卡", "概念图", "速查表", "信息图", "思维导图卡", "象限图", "商业分析图", "商业报告长图". Style covers notion subtle layered surfaces, sketch-notes hand-drawn, chalkboard, consulting-report. Layout covers dense / list / mindmap / quadrant / flow / report. Do not use for movie posters, single-character illustration, standalone photorealistic scenes, or social-feed cover series.

2026-05-144

retro-halftone-hero

inspirepan/nano-banana-playground

Create premium retro halftone hero images, landscape campaign visuals, travel-poster style website headers, and high-end promotional posters from either a fresh scene prompt or an existing photo/image reference. Use whenever the user asks for retro halftone, vintage print texture, premium travel poster, CMYK dots, ink bleed, analog print, paper texture over a normal photo, or 高级复古半调、网点纹理、风景宣传图、海报质感、旅行海报、照片加半调质感. Strong visual core is a cinematic base image plus CMYK / riso halftone dots driven by tonal values, subtle ink bleed, paper grain, and restrained color. Do NOT use for comic panels, dense infographics, UI mockups, product cutouts, screen-print duotone editorial posters with heavy typography, or cheap global filter effects.

2026-05-144

name

encyclopedia-card

description

Encyclopedia Card

为单一主题（动物、植物、食物、现象、工具、地标、概念等）生成一张竖版百科全书式图鉴卡。视觉语言是可出版的博物图鉴 + 现代编辑信息图——一张精美主视觉 + 多个局部放大标注 + 圆角模块化面板 + Top 5 / 评分 / 摘要卡片。浅色背景、柔和配色、精致小图标、高信息密度但阅读节奏平静。不是广告海报、不是爆款新闻封面、不是漫画插画。

核心流程

Agent 先确定主题、收集事实、挑选信息模块、准备每个模块的具体文本，再把这些具体内容填进一段纯视觉指令 prompt 交给 GenImage。生图模型只接收已填好的具体内容——不要保留"自动选择模块"、"根据主题自适应"这类 meta 指令。

确认主题：用户一句话表达（"给大熊猫做一张"、"做一张意式浓缩 Espresso 图鉴"）即足够；歧义时用一次 AskUserQuestion 问清方向（物种 vs 文化现象、实物 vs 概念）。
收集事实：优先使用已知世界知识；事实稀缺或需要最新数据时用 WebSearch / WebFetch，把关键数字、分类信息、原产地等落成结构化 notes。不要编造不存在的事实、数字、Top 5 排名。
按 §2 选 4-7 个信息模块，每个模块准备具体字段值（label + 2-4 条 bullet 或数值）。
按 §3 挑一组配色（通常跟主题气质挂钩——生物走森林绿或博物米黄，食物饮品走奶油暖，科技物品走石板蓝灰）。
按 §5 模板拼 prompt，把模块、评分、Top 5 等具体字符串填进去。
调一次 GenImage（默认参数见 §6）。UI 审批卡就是确认入口。
用户反馈后重跑 GenImage，只改必要字段（替换某个模块、换配色、调整评分）。

2. 信息模块选择（Agent 在拼 prompt 前完成）

准备 4-7 个模块。按主题类型从下表挑组合；不要 12 个模块全上，留白和阅读舒适度比信息密度更重要。

模块 key	说明 / 填什么	常用主题
`profile`	基本资料：别名 / 学名 / 分类 / 原产地 / 出现时期	全部主题建议首选
`taxonomy`	分类学 / 族群 / 家族树；可画成小型分级图	生物、植物、食物品类
`physical`	物理特征：尺寸 / 重量 / 颜色 / 形态；带数值	生物、物品、地标
`behavior`	行为 / 生态 / 习性 / 饮食 / 活动节律	动物、昆虫、鸟类
`structure`	结构 / 解剖 / 形成机制 / 工作原理	植物、地质现象、工具、设备
`usage`	使用方法 / 步骤 / 冲泡 / 烹饪 / 操作流程	食物饮品、工具、工艺
`growth`	生长条件 / 栽培周期 / 温湿度 / 光照	植物、作物、菌菇
`care`	护理 / 维护 / 保养 / 优化建议	宠物、植物、设备、技能
`risks`	风险 / 警告 / 副作用 / 注意事项	食物饮品、药草、野外生物
`scenarios`	适用用户 / 应用场景 / 搭配建议	工具、饮品、装备
`pros_cons`	优点 / 缺点对比；左右两列	工具、产品、选项决策
`ratings`	视觉评分条：3-5 项子指标 + 横条或 1-5 星	食物饮品、工具、产品、自然地点
`top5`	Top 5 列表：品种 / 变种 / 子分类 / 相关推荐	生物变种、食物流派、工具款式
`timeline`	时间线 / 历史重要节点 / 生命周期阶段	文化现象、历史、发展演化
`fun_facts`	3-5 条趣味冷知识	任意主题

写模块内容的原则：

每个 bullet 6-14 字；数值要有单位（cm / kg / °C / %）。
top5 / ratings 要真实可查；不要为了凑视觉编造排名。
不要所有模块都写满 4 条；让某些模块只 2-3 条，形成视觉呼吸。

3. 配色主题（挑一组锁定 5-7 色）

浅色背景 + 柔和配色是本 skill 的基本气质——绝不上深色背景、饱和霓虹、高冲击红黑。品牌 / 主题有强色时作为主色，其他跟随。

主题	背景 bg	卡片 surface	Primary 主题色	Accent 点缀	Text / Muted	适合
`cream-warm`	`#FAF6EE` 奶油米	`#FFFFFF`	`#8B5E3C` 可可棕	`#D4A574` 蜜金	`#2A2520` / `#8A8378`	食物饮品、咖啡、面包
`botanical`	`#F5F2E8` 羊皮米	`#FFFFFF`	`#3F6B3E` 植物绿	`#C97B3A` 陶土橙	`#263526` / `#7E8A7E`	植物、作物、香料、园艺
`field-guide`	`#F2EADA` 泛黄手册米	`#FFFDF7`	`#3A4A5C` 墨蓝灰	`#B85C38` 暖砖红	`#2D2A23` / `#877D6D`	鸟类、昆虫、博物学
`ocean`	`#EEF3F5` 浅雾蓝	`#FFFFFF`	`#2F5D6E` 深海蓝	`#E0A84B` 海盐金	`#1F3540` / `#7A8E96`	海洋生物、水域、天气
`slate-modern`	`#F3F4F6` 石板灰白	`#FFFFFF`	`#1F2937` 石墨黑	`#E07A3F` 橙	`#111827` / `#6B7280`	工具、科技物件、设备
`pastel-life`	`#FFF4EE` 蜜桃奶	`#FFFFFF`	`#C77D6A` 陶瓷粉	`#7BA89A` 薄荷绿	`#3A2A25` / `#9A8780`	宠物、生活方式、手账
`museum-ivory`	`#F6EFDD` 象牙米	`#FFFDF4`	`#5C3A21` 深咖	`#A47148` 驼	`#2B1F14` / `#8A7A63`	文化遗产、历史、器物

锁定 5 色：bg / surface / primary / accent / text（+ muted 作为第 6 辅助色，仅用于次级标签 / 描述文本 / 分隔线）。

4. 版式骨架

竖版（3:4 或 9:16 或 2:3）。从上到下的基础节律：

页眉 Header (占 8-10%)：小型分类标签（"BOTANICAL GUIDE" / "FIELD NOTES" / "食物百科") + 顺序编号 (NO. 03 / VOL. 01) + 装饰细线。
标题区 Title Block (10-12%)：主题大标题（中文用粗黑或粗宋；英文用 serif / display）+ 拉丁学名或副标题（斜体 serif，字号小）。
主视觉 Hero (25-35%)：主体居中或略偏一侧；2-4 个局部放大细节（圆形镜头或椭圆 plate + 细引线 + 指向箭头 + 标注文字）。
主信息模块区 Modules (35-45%)：圆角卡片网格（通常 2 列 × 2-3 行或 1 列堆叠 + 2 列混合）；每张卡带小 icon + 标题 + 2-4 条 bullet。
评分 / Top5 / 摘要条 (8-12%)：横向评分条 / Top 5 列表 / 快速摘要 chips；不要把评分和 Top 5 同时都塞进去。
页脚 Footer (3-5%)：系列名 / 日期 / 小小 "collect & share" 或作者签名线、一条装饰线。

整体留白 20-30%，感觉是一页真的会被印在手册里的纸张，而不是海报。

5. Prompt 拼装模板（交给 GenImage 的完整文本）

把下面文本作为 GenImage.prompt 内容。方括号占位全部替换为 §2 / §3 准备好的具体字符串；不要保留任何 [] 或 "根据主题自动..."、"选择合适的..." 之类的 meta 指令。字段缺失就删行，不要留空占位。

Role: Senior editorial infographic designer producing a publishable encyclopedia / field-guide card. Output a single vertical knowledge card in the aesthetic of a premium natural-history guidebook mixed with modern editorial infographic design. NOT an ad poster, NOT a social media thumbnail, NOT a dense business dashboard — it should read like a page from a collectible reference handbook that belongs in a consistent series.

# Series / page meta
- Series label (small eyebrow text at top): [series_label, e.g. "BOTANICAL FIELD GUIDE" / "COFFEE ENCYCLOPEDIA" / "生物图鉴"]
- Issue number (small, right-aligned): [issue_no, e.g. "NO. 07" / "VOL. 01 · 03"]
- Footer signature line (optional): [footer_line, e.g. "COLLECT · READ · SHARE" or a short tagline]

# Title block
- Main title (EXACT text, large, dominant, 中文用粗黑 / 粗宋, 英文用 editorial serif display): "[title verbatim]"
- Subtitle / latin name / tagline (smaller, italic serif for latin names): "[subtitle verbatim]"

# Hero illustration — one main subject
Render one beautifully detailed illustration of: [hero_subject_description, e.g. "a male ruby-throated hummingbird in 3/4 profile, hovering, wings mid-beat, realistic botanical-plate style"]
- Medium: [medium, e.g. "refined gouache + ink line" / "soft watercolor with subtle stippling" / "flat editorial vector with textured fills"]
- Placement: hero sits slightly left of center in the top third; keeps breathing room around it.
- NO text rendered inside the hero illustration itself.

# Detail callouts (zoomed-in features)
Render [N, e.g. 3] small circular or oval "lens" zoom-ins around the hero, each showing a magnified feature with a thin dotted / solid leader line pointing from the lens to the exact anatomy on the hero:

[for each callout, output a block:]
- Callout: anchor="[anatomy, e.g. 'wing primary feather']" / shown_in_lens="[what the zoom shows]" / label="[label text, 3-8 字, 中英皆可]"

Example:
- Callout: anchor="喙" / shown_in_lens="细长中空虹吸结构" / label="虹吸式长喙"

Lens style: circular frame, 1-1.5 px accent-color ring, subtle inner shadow, label sits next to the lens in small caps or 小号粗体.

# Modular info panels (rounded cards)
Render [M, 4-7] rounded info panels on a clean grid below the hero. Each panel has: small line icon (18-22 px) in the accent color, a short bold header, and 2-4 short bullet lines.

[for each panel, output a block with verbatim content:]
- Panel: icon_hint="[icon concept, e.g. 'leaf' / 'scale' / 'thermometer' / 'book']" / title="[panel_title]" / bullets=[ "[bullet 1]", "[bullet 2]", "[bullet 3]" ]

Example panels (replace with the real ones for this topic):
- Panel: icon_hint="id-card" / title="基本资料" / bullets=["学名：Trochilidae", "分布：美洲", "体长：6–13 cm", "体重：2–20 g"]
- Panel: icon_hint="ruler" / title="物理特征" / bullets=["羽色金属光泽", "翅膀每秒 50–80 次", "喙长而中空"]

Rules:
- Card shape: 8-12 px rounded corners, white or [surface hex] fill, very soft shadow (y-offset 2-4 px, blur 8-16 px, opacity 8-12%).
- Grid: prefer 2 columns × [rows] or a mix of one wide card + 2 smaller cards per row. Not a uniform grid of identical tiles.
- Bullet text: left-aligned, 13-15 px equivalent, 1.5 line-height.

# Quick-visual module (choose ONE — ratings OR top5 OR summary — do NOT stack all three)
Selected module: [ratings | top5 | summary]

If `ratings`:
- Horizontal rating bars for 3-5 sub-metrics. Each row: label + filled bar + "4.5 / 5" or stars. Render these EXACT metrics:
[for each: "[metric_label]": [score_out_of_5]]

If `top5`:
- A "TOP 5 [category]" block with numbered 01–05 rows. Each row: number badge + name + one-line descriptor.
[for each row: "[rank]. [name] — [one-line descriptor]"]

If `summary`:
- 4-6 small chip tags with one-word or 2-3 word highlights, plus one sentence "Takeaway:" line below.
[chips and takeaway sentence]

# Palette (strict 5-6 colors, do NOT render hex text)
- Background: [bg hex]
- Card surface: [surface hex]
- Primary (title, icon strokes, key numbers): [primary hex]
- Accent (lens rings, chip highlights, rating bars, small decorative dots): [accent hex]
- Text: [text hex]
- Muted (sub-labels, bullet secondary lines, footer): [muted hex]

Do not render hex values, color names, palette labels, or the words "palette" / "accent" / "module" / "callout" as visible text in the image.

# Typography rules
- Title: [title_font_hint, e.g. "粗黑 (思源黑体 Heavy) for Chinese, Playfair-style display serif for English"]
- Latin name / subtitle: italic editorial serif, smaller
- Panel headers: 中文粗黑 / bold sans
- Bullet body: regular sans, optically sized for reading
- Numbers and scores: tabular numerals
- Small caps / label eyebrows: letter-spaced 0.08–0.12em, uppercase, muted color

# Illustration + texture
- Keep the overall page clean — large areas of the background hex are visible.
- Sprinkle tiny, restrained decorative motifs (3-6 total) tied to the subject: small botanical sprigs, constellations of dots, tiny footprints, measuring-ruler marks — never busy patterns behind text.
- Very subtle paper-grain texture at 4-8% opacity over the full card is OK; no heavy noise, no halftone, no grunge.
- Shadows are soft and short, not dramatic.

# Layout rhythm
From top to bottom:
1. Eyebrow series label + issue number (thin divider line below)
2. Title block (title + subtitle)
3. Hero illustration with detail callouts
4. Info panel grid ([M] panels)
5. Quick-visual module ([ratings | top5 | summary])
6. Footer line with signature / date / collect-cue

Leave 20-30% whitespace total; reading pace should feel calm.

# Hard constraints
- Vertical aspect only (3:4 / 2:3 / 9:16); this card would be printed on a handbook page.
- All text must be legible and spelled correctly; match language to [language, zh / en / bilingual].
- Do NOT add a glossy ad-poster feel, no heavy drop shadows, no neon, no saturated gradients, no commercial CTAs.
- NO price tags, NO QR codes, NO watermarks unless the user asked.
- No realistic human faces; people appear only as stylized illustration if the topic truly needs them.
- Stay inside the 5-6 palette colors; do not introduce a 7th decorative hue.
- Hero illustration must NOT contain text inside it.
- Panels must have DISTINCT contents; do not repeat the same bullet across panels.

6. GenImage 调用

因为文本密度高（标题、学名、多面板 bullet、评分数字），优先使用 gpt-image-2——中文 + 英文学名混排最稳。如果是纯英文且以插画为主、文字极少，可换 nano-banana-pro。

{
  "image_id": "encyclopedia_<slug>",
  "prompt": "<按 §5 拼装好的完整文本>",
  "model": "gpt-image-2",
  "resolution": "2K",
  "ratio": "3:4",
  "n": 1,
  "reference_image_ids": []
}

参数	默认	备注
model	`gpt-image-2`	多语言 + 密集小字段；纯插画少字可用 `nano-banana-pro`
resolution	`2K`	想做印刷级上 `4K`
ratio	`3:4`	默认；手机分发 `9:16`；长细手册 `2:3`
n	`1`	系列对比再升到 2-3（比如"同一生物不同阶段"）
reference_image_ids	`[]`	仅当用户提供了真实的主体照片、参考插画，且其 image id 已存在时才填

7. 参考图处理

如果用户给了主体照片（比如"我家的猫，照着它画一张图鉴卡"）或历史图鉴卡，把 image id 填进 reference_image_ids，并在 prompt 末尾追加：

# Reference image directives — MUST INCORPORATE
CRITICAL: The hero illustration must visually match the reference subject.
- Identity: [distinctive markings / color patches / posture / breed]
- Stylize to [medium]; NO photoreal face or photographic shading
- Preserve [signature features user cares about]
- Color mapping: keep hero identity colors, but adjust surrounding palette to match the page palette above

8. 跟进调整

"信息太挤 / 留白更多" → 把模块数从 7 砍到 4-5，或把 Top 5 换成 3 个 summary chips。
"配色再柔和 / 再复古" → 从 §3 换一组 palette（cream-warm ↔ museum-ivory ↔ pastel-life），其它不动。
"主视觉换姿态 / 换物种" → 改 hero_subject_description，其它模块保留；image_id 改后缀 _v2。
"加一个 Top 5" → 在 Quick-visual 选 top5；如果已有 ratings，选一个保留，不要叠加。
"变横版" → 换 article-cover-image 或 knowledge-infographic，本 skill 只做竖版。
"要 9:16 手机壁纸" → 改 ratio: "9:16"；同时在模板里让标题区和主视觉比例更高（hero 占 30-40%），否则下面模块会被压扁。

9. 系列化（collect as a series）

用户常常会连续做同一系列的多张（"再做一张北极狐"、"加一张北极兔"）。保持系列感需要：

统一使用同一套 palette（首次做的 palette 是该系列的基因色）。
统一 series_label（BOTANICAL FIELD GUIDE / 深海图鉴 · VOL 1）。
保持 title typography family 和 hero medium 不变；只换主体和模块内容。
issue_no 递增（NO. 01 → NO. 02 → NO. 03）。
image_id 用系列 + 主题命名：encyclopedia_field_guide_arctic_fox / encyclopedia_field_guide_arctic_hare。

10. 常见反例（写 prompt 时主动规避）

把 §2 的模块选择指令原样塞给生图模型：绝不出现 "从下列模块中自动挑选"、"根据主题自适应" 等话术；生图模型只看到已经选好的 N 个具体模块 + 具体内容。
编造事实 / Top 5 / 评分：所有数值 / 排名 / 学名必须是真实的；不确定时用 WebSearch 查，宁可减少模块也不要写假数据。
海报化过度：出现大块深色背景、霓虹、夸张渐变、营销文案 CTA → 失去手册气质。
模块全部等大排成 2×3 网格：缺视觉节奏；应错落（1 wide + 2 narrow）并让内容长度自然决定高度。
主视觉里塞字：hero 区域里不要渲染任何文本；所有文字只出现在标题区、callout label、panel 内、footer。
callout 引线指向模糊位置：引线必须从 lens 圆圈准确指到主视觉上那个解剖位置，不能只漂在空白处。
palette 扩散到 7+ 色：装饰 sprig / dots 只用 primary + accent 两色，不要引入第 7 种装饰色。
英文学名拼错 / 语种混用混乱：拉丁学名用斜体 serif，中文学名不斜体；混排保持语种分层清晰。