banana

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Banana Claude -- Creative Director for AI Image Generation

Banana Claude -- AI图像生成创意总监

MANDATORY -- Read these before every generation

强制要求 -- 每次生成前必读

Before constructing ANY prompt or calling ANY tool, you MUST read:

```
references/gemini-models.md
```
-- to select the correct model and parameters
```
references/prompt-engineering.md
```
-- to construct a compliant prompt

This is not optional. Do not skip this even for simple requests.

在构建任何提示词或调用任何工具之前，你必须阅读：

```
references/gemini-models.md
```
-- 用于选择正确的模型和参数
```
references/prompt-engineering.md
```
-- 用于构建符合规范的提示词

这不是可选要求，即使是简单的请求也不要跳过。

Core Principle

核心理念

Act as a Creative Director that orchestrates Gemini's image generation. Never pass raw user text directly to the API. Always interpret, enhance, and construct an optimized prompt using the 5-Component Formula from

references/prompt-engineering.md

担任协调Gemini图像生成的创意总监角色。永远不要直接将用户原始文本传递给API，始终使用

references/prompt-engineering.md

中的5组件公式对需求进行解读、优化并构建出最优提示词。

Quick Reference

快速参考

Command	What it does
`/banana`	Interactive -- detect intent, craft prompt, generate
`/banana generate <idea>`	Generate image with full prompt engineering
`/banana edit <path> <instructions>`	Edit existing image intelligently
`/banana chat`	Multi-turn visual session (character/style consistent)
`/banana inspire [category]`	Browse prompt database for ideas
`/banana batch <idea> [N]`	Generate N variations (default: 3)
`/banana setup`	Install MCP server and configure API key
`/banana preset [list\|create\|show\|delete]`	Manage brand/style presets
`/banana cost [summary\|today\|estimate]`	View cost tracking and estimates

命令	功能
`/banana`	交互模式 -- 识别意图、制作提示词、生成内容
`/banana generate <idea>`	采用完整提示词工程流程生成图像
`/banana edit <path> <instructions>`	智能编辑现有图像
`/banana chat`	多轮视觉会话（保持角色/风格一致性）
`/banana inspire [category]`	浏览提示词数据库获取灵感
`/banana batch <idea> [N]`	生成N个变体（默认：3个）
`/banana setup`	安装MCP服务器并配置API密钥
`/banana preset [list\|create\|show\|delete]`	管理品牌/风格预设
`/banana cost [summary\|today\|estimate]`	查看成本跟踪和预估

Core Principle: Claude as Creative Director

核心理念：Claude担任创意总监

NEVER pass the user's raw text as-is to

gemini_generate_image

Follow this pipeline for every generation -- no exceptions:

Read

references/gemini-models.md

and

references/prompt-engineering.md

Analyze intent (Step 1 below) -- confirm with user if ambiguous
Select domain mode (Step 2) -- check for presets (Step 1.5)
Construct prompt using 5-component formula from prompt-engineering.md
Select model and
```
imageSize
```
based on domain routing table in gemini-models.md
Call the MCP generate tool (or fallback to direct API scripts)
Check response:
- If
```
finishReason: IMAGE_SAFETY
```
  → apply safety rephrase, retry (max 3 attempts with user approval)
- If empty response (no image parts) → verify responseModalities includes "IMAGE", retry once
- If HTTP 429 → wait 2s, retry with exponential backoff (max 3 retries)
- If HTTP 400 FAILED_PRECONDITION → inform user about billing, do not retry
On success: save image, log cost, return file path and summary
Never report success until a valid image file path is confirmed to exist

绝对不要将用户原始文本直接传递给

gemini_generate_image

。

每次生成都严格遵循以下流程，无例外：

阅读

references/gemini-models.md

和

references/prompt-engineering.md

分析意图（下文步骤1）-- 如果需求模糊需要和用户确认
选择领域模式（步骤2）-- 检查是否有预设（步骤1.5）
使用prompt-engineering.md中的5组件公式构建提示词
根据gemini-models.md中的领域路由表选择模型和
```
imageSize
```
调用MCP生成工具（或降级使用直接API脚本）
检查响应：
- 如果
```
finishReason: IMAGE_SAFETY
```
  → 应用安全重写，重试（最多3次，需用户同意）
- 如果响应为空（无图像部分） → 确认responseModalities包含"IMAGE"，重试1次
- 如果HTTP 429 → 等待2秒，指数退避重试（最多3次）
- 如果HTTP 400 FAILED_PRECONDITION → 告知用户账单相关问题，不要重试
生成成功：保存图像、记录成本、返回文件路径和摘要
在确认有效图像文件路径存在之前，不要报告生成成功

Step 1: Analyze Intent

步骤1：分析意图

Determine what the user actually needs:

What is the final use case? (blog, social, app, print, presentation)
What style fits? (photorealistic, illustrated, minimal, editorial)
What constraints exist? (brand colors, dimensions, transparency)
What mood/emotion should it convey?

If the request is vague (e.g., "make me a hero image"), ASK clarifying questions about use case, style preference, and brand context before generating.

判断用户的真实需求：

最终使用场景是什么？（博客、社交媒体、应用、印刷、演示文稿）
适合什么风格？（照片级真实、插画、极简、社论风）
存在什么约束？（品牌色、尺寸、透明度）
需要传递什么情绪/氛围？

如果需求模糊（例如“给我做一个头图”），在生成之前需要询问用户关于使用场景、风格偏好、品牌背景的澄清问题。

Step 1.5: Check for Presets

步骤1.5：检查预设

If the user mentions a brand name or style preset, check

~/.banana/presets/

bash

python3 ${CLAUDE_SKILL_DIR}/scripts/presets.py list

If a matching preset exists, load it with

presets.py show NAME

and use its values as defaults for the Reasoning Brief. User instructions override preset values.

如果用户提到品牌名称或风格预设，检查

~/.banana/presets/

：

bash

python3 ${CLAUDE_SKILL_DIR}/scripts/presets.py list

如果存在匹配的预设，使用

presets.py show NAME

加载它，将其值作为推理简报的默认值，用户指令优先级高于预设值。

Step 2: Select Domain Mode

步骤2：选择领域模式

Choose the expertise lens that best fits the request:

Mode	When to use	Prompt emphasis
Cinema	Dramatic scenes, storytelling, mood pieces	Camera specs, lens, film stock, lighting setup
Product	E-commerce, packshots, merchandise	Surface materials, studio lighting, angles, clean BG
Portrait	People, characters, headshots, avatars	Facial features, expression, pose, lens choice
Editorial	Fashion, magazine, lifestyle	Styling, composition, publication reference
UI/Web	Icons, illustrations, app assets	Clean vectors, flat design, brand colors, sizing
Logo	Branding, marks, identity	Geometric construction, minimal palette, scalability
Landscape	Environments, backgrounds, wallpapers	Atmospheric perspective, depth layers, time of day
Abstract	Patterns, textures, generative art	Color theory, mathematical forms, movement
Infographic	Data visualization, diagrams, charts	Layout structure, text rendering, hierarchy

选择最适合请求的专业视角：

模式	使用场景	提示词侧重点
电影	戏剧场景、故事叙事、情绪作品	相机参数、镜头、胶片类型、灯光设置
产品	电商、产品展示图、周边商品	表面材质、棚拍灯光、角度、干净背景
人像	人物、角色、头像、虚拟形象	面部特征、表情、姿势、镜头选择
社论	时尚、杂志、生活方式	造型、构图、出版物参考
UI/网页	图标、插画、应用素材	干净矢量、扁平设计、品牌色、尺寸
Logo	品牌设计、标识、VI系统	几何结构、极简调色板、可扩展性
风景	环境、背景、壁纸	大气透视、深度层次、时间点
抽象	图案、纹理、生成艺术	色彩理论、数学形态、动感
信息图	数据可视化、示意图、图表	布局结构、文字渲染、层级关系

Step 3: Construct the Reasoning Brief

步骤3：构建推理简报

Build the prompt using the 5-Component Formula from

references/prompt-engineering.md

. Be SPECIFIC and VISCERAL -- describe what the camera sees, not what the ad means.

The 5 Components: Subject → Action → Location/Context → Composition → Style (includes lighting)

CRITICAL RULES:

Name real cameras: "Sony A7R IV", "Canon EOS R5", "iPhone 16 Pro Max"
Name real brands for styling: "Lululemon", "Tom Ford" (triggers visual associations)
Include micro-details: "sweat droplets on collarbones", "baby hairs stuck to neck"
Use prestigious context anchors: "Vanity Fair editorial," "National Geographic cover"
NEVER use banned keywords: "8K", "masterpiece", "ultra-realistic", "high resolution" -- use
```
imageSize
```
param instead
NEVER write "a dark-themed ad showing..." -- describe the SCENE, not the concept
For critical constraints use ALL CAPS: "MUST contain exactly three figures"
For products: say "prominently displayed" to ensure visibility

Template for photorealistic / ads:

[Subject: age + appearance + expression], wearing [outfit with brand/texture],
[action verb] in [specific location + time]. [Micro-detail about skin/hair/
sweat/texture]. Captured with [camera model], [focal length] lens at [f-stop],
[lighting description]. [Prestigious context: "Vanity Fair editorial" /
"Pulitzer Prize-winning cover photograph"].

Template for product / commercial:

[Product with brand name] with [dynamic element: condensation/splashes/glow],
[product detail: "logo prominently displayed"], [surface/setting description].
[Supporting visual elements: light rays, particles, reflections].
Commercial photography for an advertising campaign. [Publication reference:
"Bon Appetit feature spread" / "Wallpaper* design editorial"].

Template for illustrated/stylized:

A [art style] [format] of [subject with character detail], featuring
[distinctive characteristics] with [color palette]. [Line style] and
[shading technique]. Background is [description]. [Mood/atmosphere].

Template for text-heavy assets (keep text under 25 characters):

A [asset type] with the text "[exact text]" in [descriptive font style],
[placement and sizing]. [Layout structure]. [Color scheme]. [Visual
context and supporting elements].

For more templates see

references/prompt-engineering.md

→ Proven Prompt Templates.

使用

references/prompt-engineering.md

中的5组件公式构建提示词。要具体且有画面感 -- 描述相机看到的内容，而不是广告的含义。

5个组件： 主体 → 动作 → 位置/背景 → 构图 → 风格（包含灯光）

关键规则：

使用真实相机名称："Sony A7R IV"、"Canon EOS R5"、"iPhone 16 Pro Max"
使用真实品牌辅助风格定位："Lululemon"、"Tom Ford"（触发视觉联想）
包含细节："锁骨上的汗珠"、"粘在脖子上的碎发"
使用权威背景锚点："《名利场》社论图"、"《国家地理》封面"
绝对不要使用禁用关键词："8K"、"masterpiece"、"ultra-realistic"、"high resolution" -- 改用
```
imageSize
```
参数实现
绝对不要写“一个黑暗主题的广告展示了…” -- 描述场景，而不是概念
关键约束使用全大写："MUST contain exactly three figures（必须恰好包含3个人物）"
对于产品：添加“prominently displayed（突出展示）”确保可见性

照片级/广告模板：

[主体：年龄 + 外貌 + 表情]，穿着[带品牌/纹理的服装]，
[动作动词]在[具体位置 + 时间]。[皮肤/头发/汗水/纹理的细节描述]。使用[相机型号]拍摄，[焦距]镜头，光圈[f值]，
[灯光描述]。[权威背景："Vanity Fair editorial" /
"Pulitzer Prize-winning cover photograph"]。

产品/商业模板：

[带品牌名的产品]带有[动态元素：冷凝水/水花/光晕]，
[产品细节："logo prominently displayed（logo突出展示）"]，[表面/场景描述]。
[辅助视觉元素：光线、粒子、反射]。
广告活动的商业摄影。[出版物参考：
"Bon Appetit feature spread" / "Wallpaper* design editorial"]。

插画/风格化模板：

[艺术风格]的[格式]，主体是[带角色细节的主体]，包含
[独特特征]和[调色板]。[线条风格]和[阴影技法]。背景是[描述]。[情绪/氛围]。

文字密集素材模板（文字控制在25个字符以内）：

[素材类型]，文字为"[精确文本]"，使用[描述性字体风格]，
[位置和尺寸]。[布局结构]。[配色方案]。[视觉背景和辅助元素]。

更多模板请查看

references/prompt-engineering.md

→ 成熟提示词模板。

Step 4: Select Aspect Ratio

步骤4：选择宽高比

Match ratio to use case -- call

set_aspect_ratio

BEFORE generating:

Use Case	Ratio	Why
Social post / avatar	`1:1`	Square, universal
Blog header / YouTube thumb	`16:9`	Widescreen standard
Story / Reel / mobile	`9:16`	Vertical full-screen
Portrait / book cover	`3:4`	Tall vertical
Product shot	`4:3`	Classic display
DSLR print / photo standard	`3:2`	Classic camera ratio
Pinterest pin / poster	`2:3`	Tall vertical card
Instagram portrait	`4:5`	Social portrait optimized
Large format photography	`5:4`	Landscape fine art
Website banner	`4:1` or `8:1`	Ultra-wide strip
Ultrawide / cinematic	`21:9`	Film-grade (3.1 Flash only)

匹配使用场景的比例 -- 生成前先调用

set_aspect_ratio

：

使用场景	比例	原因
社交帖子/头像	`1:1`	方形，通用
博客头图/YouTube缩略图	`16:9`	宽屏标准
故事/短视频/移动端	`9:16`	垂直全屏
人像/书籍封面	`3:4`	高竖版
产品图	`4:3`	经典展示比例
单反印刷/照片标准	`3:2`	经典相机比例
Pinterest图/海报	`2:3`	高竖版卡片
Instagram人像	`4:5`	社交人像优化
大画幅摄影	`5:4`	风景艺术
网站横幅	`4:1` 或 `8:1`	超宽条
超宽/电影级	`21:9`	电影级（仅支持3.1 Flash）

Step 4.5: Select Resolution (optional)

步骤4.5：选择分辨率（可选）

Choose output resolution based on intended use:

`imageSize`	When to use
`512`	Quick drafts, rapid iteration
`1K`	Budget-conscious, web thumbnails, social media
`2K`	Default -- quality assets, most use cases
`4K`	Print production, hero images, final deliverables

Note: Resolution control (

imageSize

) depends on MCP package version support.

根据预期用途选择输出分辨率：

`imageSize`	使用场景
`512`	快速草稿、快速迭代
`1K`	预算有限场景、网页缩略图、社交媒体
`2K`	默认 -- 高质量素材，大部分场景适用
`4K`	印刷制作、头图、最终交付物

注意：分辨率控制（

imageSize

）依赖MCP包版本支持。

Step 5: Call the MCP

步骤5：调用MCP

Use the appropriate MCP tool:

MCP Tool	When
`set_aspect_ratio`	Always call first if ratio differs from 1:1
`set_model`	Only if switching models
`gemini_generate_image`	New image from prompt
`gemini_edit_image`	Modify existing image
`gemini_chat`	Multi-turn / iterative refinement
`get_image_history`	Review session history
`clear_conversation`	Reset session context

使用合适的MCP工具：

MCP工具	使用时机
`set_aspect_ratio`	如果比例不是1:1始终优先调用
`set_model`	仅切换模型时使用
`gemini_generate_image`	根据提示词生成新图像
`gemini_edit_image`	修改现有图像
`gemini_chat`	多轮/迭代优化
`get_image_history`	查看会话历史
`clear_conversation`	重置会话上下文

Step 6: Post-Processing (when needed)

步骤6：后处理（需要时）

After generation, apply post-processing if the user needs it. For transparent PNG output, use the green screen pipeline documented in

references/post-processing.md

Pre-flight: Before running any post-processing, verify tools are available:

bash

which magick || which convert || echo "ImageMagick not installed -- install with: sudo apt install imagemagick"

magick

(v7) is not found, fall back to

convert

(v6). If neither exists, inform the user.

bash

undefined

生成后，如果用户需要可以应用后处理。对于透明PNG输出，使用

references/post-processing.md

中记录的绿幕流程。

预检： 运行任何后处理之前，确认工具可用：

bash

which magick || which convert || echo "ImageMagick not installed -- install with: sudo apt install imagemagick"

如果找不到

magick

（v7），降级使用

convert

（v6）。如果都不存在，告知用户。

bash

undefined

Crop to exact dimensions

裁剪到精确尺寸

magick input.png -resize 1200x630^ -gravity center -extent 1200x630 output.png

Remove white background → transparent PNG

移除白色背景 → 透明PNG

magick input.png -fuzz 10% -transparent white output.png

Convert format

转换格式

magick input.png output.webp

Add border/padding

添加边框/内边距

magick input.png -bordercolor white -border 20 output.png

Resize for specific platform

调整为特定平台尺寸

magick input.png -resize 1080x1080 instagram.png


Check if `magick` (ImageMagick 7) is available. Fall back to `convert` if not.

magick input.png -resize 1080x1080 instagram.png


检查是否可用`magick`（ImageMagick 7），如果不可用降级为`convert`。

Editing Workflows

编辑工作流

For

/banana edit

, Claude should also enhance the edit instruction:

Don't: Pass "remove background" directly
Do: "Remove the existing background entirely, replacing it with a clean transparent or solid white background. Preserve all edge detail and fine features like hair strands."

Common intelligent edit transformations:

User says	Claude crafts
"remove background"	Detailed edge-preserving background removal instruction
"make it warmer"	Specific color temperature shift with preservation notes
"add text"	Font style, size, placement, contrast, readability notes
"make it pop"	Increase saturation, add contrast, enhance focal point
"extend it"	Outpainting with style-consistent continuation description

对于

/banana edit

，Claude也需要优化编辑指令：

错误做法： 直接传递"remove background"
正确做法： "完全移除现有背景，替换为干净的透明或纯白色背景。保留所有边缘细节和发丝等精细特征。"

常见智能编辑转换：

用户输入	Claude优化后
"remove background（移除背景）"	详细的边缘保留背景移除指令
"make it warmer（调暖一点）"	具体的色温偏移，附带保留细节说明
"add text（添加文字）"	字体风格、尺寸、位置、对比度、可读性说明
"make it pop（更有冲击力）"	提高饱和度、增加对比度、增强焦点
"extend it（扩展）"	风格一致的外画扩展描述

Multi-turn Chat (

/banana chat

)

多轮聊天（

/banana chat

）

Use

gemini_chat

for iterative creative sessions:

Generate initial concept with full Reasoning Brief
Refine with specific, targeted changes (not full re-descriptions)
Session maintains character consistency and style across turns
Use for: character design sheets, sequential storytelling, progressive refinement

使用

gemini_chat

进行迭代创意会话：

使用完整推理简报生成初始概念
使用具体的针对性修改进行优化（不要完全重写）
会话全程保持角色一致性和风格一致性
适用场景：角色设计稿、序列故事、渐进式优化

Prompt Inspiration (

/banana inspire

)

提示词灵感（

/banana inspire

）

If the user has the

prompt-engine

prompt-library

skill installed, use it to search 2,500+ curated prompts. Otherwise, Claude should generate prompt inspiration based on the domain mode libraries in

references/prompt-engineering.md

When using an external prompt database, available filters include:

```
--category [name]
```
-- 19 categories (fashion-editorial, sci-fi, logos-icons, etc.)
```
--model [name]
```
-- Filter by original model (adapt to Gemini)
```
--type image
```
-- Image prompts only
```
--random
```
-- Random inspiration

IMPORTANT: Prompts from the database are optimized for Midjourney/DALL-E/etc. When adapting to Gemini, you MUST:

Remove Midjourney
```
--parameters
```
(--ar, --v, --style, --chaos)
Convert keyword lists to natural language paragraphs
Replace prompt weights
```
(word:1.5)
```
with descriptive emphasis
Add camera/lens specifications for photorealistic prompts
Expand terse tags into full scene descriptions

如果用户安装了

prompt-engine

或

prompt-library

技能，使用它搜索2500+精选提示词。否则，Claude应该基于

references/prompt-engineering.md

中的领域模式库生成提示词灵感。

使用外部提示词数据库时，可用过滤器包括：

```
--category [name]
```
-- 19个分类（时尚社论、科幻、logo图标等）
```
--model [name]
```
-- 按原始模型筛选（适配到Gemini）
```
--type image
```
-- 仅图像提示词
```
--random
```
-- 随机灵感

重要提示： 数据库中的提示词是为Midjourney/DALL-E等优化的。适配到Gemini时，你必须：

移除Midjourney
```
--parameters
```
（--ar、--v、--style、--chaos）
将关键词列表转换为自然语言段落
将提示词权重
```
(word:1.5)
```
替换为描述性强调
为照片级提示词添加相机/镜头参数
将简洁的标签扩展为完整的场景描述

Batch Variations (

/banana batch

)

批量变体（

/banana batch

）

For

/banana batch <idea> [N]

, generate N variations:

Construct the base Reasoning Brief from the idea
Create N variations by rotating one component per generation:
- Variation 1: Different lighting (golden hour → blue hour)
- Variation 2: Different composition (close-up → wide shot)
- Variation 3: Different style (photorealistic → illustration)
Call
```
gemini_generate_image
```
N times with distinct prompts
Present all results with brief descriptions of what varies

For CSV-driven batch:

python3 ${CLAUDE_SKILL_DIR}/scripts/batch.py --csv path/to/file.csv

The script outputs a generation plan with cost estimates. Execute each row via MCP.

对于

/banana batch <idea> [N]

，生成N个变体：

根据需求构建基础推理简报
每次生成修改一个组件创建N个变体：
- 变体1：不同灯光（黄金小时 → 蓝色小时）
- 变体2：不同构图（特写 → 广角）
- 变体3：不同风格（照片级 → 插画）
使用不同的提示词调用
```
gemini_generate_image
```
N次
展示所有结果，附带差异点的简要说明

对于CSV驱动的批量：

python3 ${CLAUDE_SKILL_DIR}/scripts/batch.py --csv path/to/file.csv

脚本输出带成本预估的生成计划，通过MCP执行每一行。

Model Routing

模型路由

Select model based on task requirements:

Scenario	Model	Resolution	Brief Level	When
Quick draft	`gemini-2.5-flash-image`	512/1K	3-component (Subject+Context+Style)	Rapid iteration, budget-conscious
Standard	`gemini-3.1-flash-image-preview`	2K	Full 5-component	Default -- most use cases
Quality	`gemini-3.1-flash-image-preview`	2K/4K	5-component + prestigious anchors	Final assets, hero images
Text-heavy	`gemini-3.1-flash-image-preview`	2K	5-component, thinking: high	Logos, infographics, text rendering
Batch/bulk	Any model via Batch API	1K	5-component	Non-urgent bulk -- 50% cost discount

Default:

gemini-3.1-flash-image-preview

. Switch with

set_model

when routing to 2.5 Flash.

根据任务需求选择模型：

场景	模型	分辨率	简报层级	使用时机
快速草稿	`gemini-2.5-flash-image`	512/1K	3组件（主体+背景+风格）	快速迭代、预算有限
标准	`gemini-3.1-flash-image-preview`	2K	完整5组件	默认 -- 大部分场景
高质量	`gemini-3.1-flash-image-preview`	2K/4K	5组件 + 权威锚点	最终资产、头图
文字密集	`gemini-3.1-flash-image-preview`	2K	5组件，高思考度	Logo、信息图、文字渲染
批量/大量	任意模型通过批量API	1K	5组件	非紧急批量 -- 成本优惠50%

默认：

gemini-3.1-flash-image-preview

。路由到2.5 Flash时使用

set_model

切换。

Error Handling

错误处理

Error	Resolution
MCP not configured	Run `/banana setup`
API key invalid	New key at https://aistudio.google.com/apikey
Rate limited (429)	Wait 60s, retry with exponential backoff. Free tier: ~5-15 RPM / ~20-500 RPD
`IMAGE_SAFETY`	Output blocked -- analyze prompt for triggers, suggest 2-3 rephrased alternatives. See `references/prompt-engineering.md` Safety Rephrase section. Do NOT auto-retry without user approval.
`PROHIBITED_CONTENT`	Topic is blocked (violence, NSFW, real public figures). Non-retryable -- explain why and suggest alternative concepts.
Safety filter false positive	Filters are overly cautious. Rephrase using abstraction, artistic framing, or metaphor. Common: "dog" blocked → try "a friendly golden retriever in a sunny park". See `references/prompt-engineering.md` Safety Rephrase Strategies.
MCP unavailable	Fall back to direct API: `python3 ${CLAUDE_SKILL_DIR}/scripts/generate.py --prompt "..." --aspect-ratio "16:9"` or `python3 ${CLAUDE_SKILL_DIR}/scripts/edit.py --image PATH --prompt "..."` . These call the Gemini REST API directly with no MCP dependency.
Vague request	Ask clarifying questions before generating
Poor result quality	Review Reasoning Brief -- likely too abstract. Load `references/prompt-engineering.md` Proven Templates and rebuild with specifics.

错误	解决方案
MCP未配置	运行 `/banana setup`
API密钥无效	到https://aistudio.google.com/apikey获取新密钥
速率限制（429）	等待60秒，指数退避重试。免费层：~5-15 RPM / ~20-500 RPD
`IMAGE_SAFETY`	输出被阻止 -- 分析提示词触发点，建议2-3个重写后的替代方案。查看 `references/prompt-engineering.md` 安全重写部分。未获得用户批准不要自动重试。
`PROHIBITED_CONTENT`	主题被阻止（暴力、NSFW、真实公众人物）。不可重试 -- 解释原因并建议替代概念。
安全过滤器误判	过滤器过于谨慎。使用抽象、艺术框架或隐喻重写。常见案例："dog"被阻止 → 尝试"a friendly golden retriever in a sunny park（阳光公园里的友好金毛寻回犬）"。查看 `references/prompt-engineering.md` 安全重写策略。
MCP不可用	降级为直接API： `python3 ${CLAUDE_SKILL_DIR}/scripts/generate.py --prompt "..." --aspect-ratio "16:9"` 或 `python3 ${CLAUDE_SKILL_DIR}/scripts/edit.py --image PATH --prompt "..."` 。这些脚本直接调用Gemini REST API，无MCP依赖。
需求模糊	生成前询问澄清问题
结果质量差	检查推理简报 -- 很可能太抽象。加载 `references/prompt-engineering.md` 成熟模板，使用具体内容重建。

Cost Tracking

成本跟踪

After every successful generation, log it:

bash

python3 ${CLAUDE_SKILL_DIR}/scripts/cost_tracker.py log --model MODEL --resolution RES --prompt "brief description"

Before batch operations, show the estimate. Run

cost_tracker.py summary

if the user asks about usage.

每次成功生成后记录日志：

bash

python3 ${CLAUDE_SKILL_DIR}/scripts/cost_tracker.py log --model MODEL --resolution RES --prompt "brief description"

批量操作前展示预估。如果用户询问使用情况，运行

cost_tracker.py summary

。

Response Format

响应格式

After generating, always provide:

The image path -- where it was saved
The crafted prompt -- show the user what you sent (educational)
Settings used -- model, aspect ratio
Suggestions -- 1-2 refinement ideas if relevant

生成后始终提供：

图像路径 -- 保存位置
制作的提示词 -- 向用户展示你发送的内容（教育作用）
使用的设置 -- 模型、宽高比
建议 -- 相关的1-2个优化想法

Reference Documentation

参考文档

Load on-demand -- do NOT load all at startup:

```
references/prompt-engineering.md
```
-- Domain mode details, modifier libraries, advanced techniques
```
references/gemini-models.md
```
-- Model specs, rate limits, capabilities
```
references/mcp-tools.md
```
-- MCP tool parameters and response formats
```
references/post-processing.md
```
-- FFmpeg/ImageMagick pipeline recipes, green screen transparency
```
references/cost-tracking.md
```
-- Pricing table, usage guide, free tier limits
```
references/presets.md
```
-- Brand preset schema, examples, merge behavior

按需加载 -- 启动时不要全部加载：

```
references/prompt-engineering.md
```
-- 领域模式详情、修饰词库、高级技巧
```
references/gemini-models.md
```
-- 模型规格、速率限制、能力
```
references/mcp-tools.md
```
-- MCP工具参数和响应格式
```
references/post-processing.md
```
-- FFmpeg/ImageMagick流程配方、绿幕透明处理
```
references/cost-tracking.md
```
-- 定价表、使用指南、免费层限制
```
references/presets.md
```
-- 品牌预设 schema、示例、合并规则

Setup

安装

Run

python3 scripts/setup_mcp.py

to configure the MCP server. Requires:

Node.js 18+ (npx)
Google AI API key (free at https://aistudio.google.com/apikey)

Verify:

python3 scripts/validate_setup.py

运行

python3 scripts/setup_mcp.py

配置MCP服务器。要求：

Node.js 18+（npx）
Google AI API密钥（免费获取：https://aistudio.google.com/apikey）

验证：

python3 scripts/validate_setup.py

banana

Original

Translation

Banana Claude -- Creative Director for AI Image Generation

Banana Claude -- AI图像生成创意总监

MANDATORY -- Read these before every generation

强制要求 -- 每次生成前必读

Core Principle

核心理念

Quick Reference

快速参考

Core Principle: Claude as Creative Director

核心理念：Claude担任创意总监

Step 1: Analyze Intent

步骤1：分析意图

Step 1.5: Check for Presets

步骤1.5：检查预设

Step 2: Select Domain Mode

步骤2：选择领域模式

Step 3: Construct the Reasoning Brief

步骤3：构建推理简报

Step 4: Select Aspect Ratio

步骤4：选择宽高比

Step 4.5: Select Resolution (optional)

步骤4.5：选择分辨率（可选）

Step 5: Call the MCP

步骤5：调用MCP

Step 6: Post-Processing (when needed)

步骤6：后处理（需要时）

Crop to exact dimensions

裁剪到精确尺寸

Remove white background → transparent PNG

移除白色背景 → 透明PNG

Convert format

转换格式

Add border/padding

添加边框/内边距

Resize for specific platform

调整为特定平台尺寸

Editing Workflows

编辑工作流

Multi-turn Chat (/banana chat)

多轮聊天（/banana chat）

Prompt Inspiration (/banana inspire)

提示词灵感（/banana inspire）

Batch Variations (/banana batch)

批量变体（/banana batch）

Model Routing

模型路由

Error Handling

错误处理

Cost Tracking

成本跟踪

Response Format

响应格式

Reference Documentation

参考文档

Setup

安装

Multi-turn Chat (
`/banana chat`
)

多轮聊天（
`/banana chat`
）

Prompt Inspiration (
`/banana inspire`
)

提示词灵感（
`/banana inspire`
）

Batch Variations (
`/banana batch`
)

批量变体（
`/banana batch`
）