跳到正文
cccskills
免费在 GitHub 公开

generate-image

Generate images using nano-banana-2 MCP tools with best-practice prompting. Use this skill when the user asks to create, generate, or edit images.

中文简介正在准备中。当前显示原文说明。

集成工具Gemini
查看安装方法

包含的文件(1)

  • SKILL.md3.5 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Image Generation with Nano Banana 2

Use the generate_image, edit_image, and continue_editing MCP tools from the nano-banana-2 server.

First-Time Setup

Before generating images, verify the API key is configured:

  1. Call get_configuration_status to check if GEMINI_API_KEY is set.
  2. If the key is missing, instruct the user to add it to their MCP server environment configuration:
    • In Claude Code settings or .claude/settings.json, add GEMINI_API_KEY to the server's env block.
    • The key can be obtained from Google AI Studio.

Prompting Best Practices

Write narrative paragraphs, NOT comma-separated keyword lists. Narrative prompts achieve ~94% coherence vs ~61% for keyword lists.

  1. Start with image type: "Create an educational diagram showing...", "Generate a photorealistic photograph of...", "Design a flat-style icon depicting..."
  2. Keep text in images short (<25 characters). Quote text exactly and describe font style.
  3. Specify layout explicitly: side-by-side, top-to-bottom steps, centered with border, etc.
  4. Skip quality boosters like "4k masterpiece", "highly detailed", "award-winning" — they add noise, not quality.
  5. Be specific about what you want, not what you don't want. Positive descriptions work better than negations.

Defaults

  • resolution: "1K" — good balance of quality and speed. Use "2K" or "4K" for print or detail-heavy images.
  • aspectRatio: "1:1" — change to "16:9" for landscapes/banners, "9:16" for mobile/portrait.
  • thinking: "minimal" — use "high" for complex scenes with spatial relationships or text rendering.
  • returnInlineImage: Consider setting to false in Claude Code to avoid context window overflow. The image is still saved to disk and can be viewed by the user.
  • outputMimeType: "image/png" for quality, "image/jpeg" for smaller files.
  • numberOfImages: 1 — use 2-4 when exploring variations.

Workflow

  1. Generate: Use generate_image with a well-crafted narrative prompt.
  2. Review: Check the saved file path in the response.
  3. Refine: Use continue_editing to make adjustments. Be specific about what to change.
  4. Iterate: Each continue_editing call builds on the previous result.

Style Guidance

  • Diagrams: Specify colors, label positions, arrow directions. Use thinking: "high" for complex layouts.
  • Illustrations: Describe art style (flat, watercolor, line art), mood, and lighting.
  • Infographics: Use numbered lists cautiously (>8 sequential items are unreliable). Numbers-only approach works best for long lists.
  • Photos: Describe camera angle, lighting conditions, depth of field, and subject positioning.

Context Window Management

When generating images in Claude Code, set returnInlineImage: false to prevent base64 image data from filling the context window. The image is still saved to disk — just tell the user where to find it.

🔗 Связи

  • [[MOC - MCP Servers]] — mcp-servers
  • [[000 - Map of Maps]] — Map of Maps

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

1password

免费

Set up and use 1Password CLI (op). Use when installing the CLI, enabling desktop app integration, signing in, and reading/injecting secrets for commands.

中文简介正在准备中。当前显示原文说明。

ibragimov-oasis/vibe-coder32026年6月24日 更新

Use this skill when the user requests to review, analyze, critique, or summarize academic papers, research articles, preprints, or scientific publications. Supports comprehensive structured reviews covering methodology assessment, contribution evaluation, literature positioning, and constructive feedback generation. Trigger on queries involving paper URLs, uploaded PDFs, arXiv links, or requests like "review this paper", "analyze this research", "summarize this study", or "write a peer review".

中文简介正在准备中。当前显示原文说明。

ibragimov-oasis/vibe-coder32026年6月24日 更新

Add descriptions for new models from the HuggingFace router to chat-ui configuration. Use when new models are released on the router and need descriptions added to prod.yaml and dev.yaml. Triggers on requests like "add new model descriptions", "update models from router", "sync models", or when explicitly invoking /add-model-descriptions.

中文简介正在准备中。当前显示原文说明。

ibragimov-oasis/vibe-coder32026年6月24日 更新

Automates browser interactions for web testing, form filling, screenshots, and data extraction. Use when the user needs to navigate websites, interact with web pages, fill forms, take screenshots, test web applications, or extract information from web pages.

中文简介正在准备中。当前显示原文说明。

ibragimov-oasis/vibe-coder32026年6月24日 更新

Master advanced AgentDB features including QUIC synchronization, multi-database management, custom distance metrics, hybrid search, and distributed systems integration. Use when building distributed AI systems, multi-agent coordination, or advanced vector search applications.

中文简介正在准备中。当前显示原文说明。

ibragimov-oasis/vibe-coder32026年6月24日 更新

Create and train AI learning plugins with AgentDB's 9 reinforcement learning algorithms. Includes Decision Transformer, Q-Learning, SARSA, Actor-Critic, and more. Use when building self-learning agents, implementing RL, or optimizing agent behavior through experience.

中文简介正在准备中。当前显示原文说明。

ibragimov-oasis/vibe-coder32026年6月24日 更新

查看 ibragimov-oasis 的全部技能

报告此技能的问题