원클릭으로 Manus에서 모든 스킬 실행

gemini-nanobanana

스타252

포크52

업데이트2026년 3월 31일 15:44

Use this skill when users ask to generate, edit, or compose images with Nano Banana (Gemini image models), including text-to-image, image editing, multi-image composition, grounding, and output sizing/saving controls.

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

duotify

duotify/GitHubClawToolkit

GitHub 저장소 열기 Creator 저장소 보기

다운로드

Manus에서 실행

Do this first

Use a Node.js wrapper (@google/genai) as the primary flow (multi-turn edits, grounding, advanced generationConfig).
Use scripts/gemini-nanobanana-cli.js for quick single-turn generation/editing runs.
Do not teach raw curl; keep guidance in JS CLI/wrapper form.

Enforce these defaults

API key source: NANOBANANA_GEMINI_API_KEY with fallback GEMINI_API_KEY.
Default model: gemini-3.1-flash-image-preview (allow env override via NANOBANANA_MODEL).
Reference images: support up to 14 total.
Thinking strength: configurable (minimal|low|medium|high), default High.
Aspect ratio: default Auto.
Resolution: default 1K.
Output mode: default Images only.
Google Search grounding tool: default Disabled; enable with --google-search when needed.
Output directory: default nanobanana-output/, unless the prompt explicitly asks for another location.
512px rule: send imageConfig.imageSize as string "512" in API calls (never numeric 512).

Read references intentionally

Start in references\image-generation-api.md for operational payload rules, model behavior, sizing, thinking, grounding, and limits.
Use references\sources.md to verify source provenance and jump to upstream docs.
For canonical API behavior, read:

Output formatting rules

After successful generation, always report results in this format:
1. A short completion message (e.g. ✅ 圖片已產出)
2. The relative path to each generated file, formatted as a markdown image embed: ![description](relative/path/to/image.jpg)
3. File metadata: format (JPEG/PNG), dimensions, file size

Example final answer:

✅ 圖片已產出

![一杯抹茶拿鐵](workspaces/issue-2/nanobanana-output/matcha-latte-01.jpg)

- 格式：JPEG · 1408×768 · 757 KB

This ensures downstream systems (GitHub Issue comments, Telegram relay) can display or link to the image.

Execution pattern

Default to Node.js wrapper flows for regular usage, especially when payload control is needed.
Quick path (agent runs from workspaces/issue-N/; resolve the repo root first):
- REPO_ROOT=$(git rev-parse --show-toplevel 2>/dev/null || echo "../..") then node "$REPO_ROOT/.agents/skills/gemini-nanobanana/scripts/gemini-nanobanana-cli.js" --prompt "..."
- Add references via repeated -i/--image (up to 14).
- Enable grounding via --google-search when prompt needs fresh web context.
The API key (GEMINI_API_KEY / NANOBANANA_GEMINI_API_KEY) is injected by the workflow environment; do not hardcode it. The CLI reads it automatically from the environment.

이 저장소의 다른 Skills

같은 저장소

pr-creator

duotify/GitHubClawToolkit

Use this skill when asked to create a pull request (PR). It creates a new branch, commits with detailed Traditional Chinese (zh-TW) messages, enriches the related issue with technical context and problem definition, then creates a PR with a zh-TW description.

2026-04-29252

gemini-lyria-3

duotify/GitHubClawToolkit

Use this skill when the user wants to generate music or audio tracks from lyrics or text prompts using Gemini Lyria-3, including AI music composition, audio generation, track creation, and requests like "幫我作一首歌", "生成音樂", "根據這段歌詞產生音軌", or "compose a track".

2026-04-24252

telegram-notify

duotify/GitHubClawToolkit

Send Telegram Bot notifications for development events, CI/CD pipeline status, and AGENTS.md progress updates. Use when integrating Telegram messaging into GitHub Actions workflows, sending pipeline end summaries, notifying teams about automation progress or repository changes. Requires TELEGRAM_NOTIFY_BOT_TOKEN environment variable or GitHub Secret. Target Chat IDs are resolved from the CSV recipients list defined in AGENTS.md.

2026-04-08252

felo-superagent

duotify/GitHubClawToolkit

Felo SuperAgent API: AI conversation with real-time SSE streaming on a persistent LiveDoc canvas. Use when users want SuperAgent chat, continuous conversation, logo/branding design, or e-commerce product images. CRITICAL: When user provides ANY image/product description (e.g. '生成一張...產品照片', '我的提示詞為: 生成...', 'generate a product photo'), ALWAYS call the API to generate the actual image — NEVER rewrite the prompt as text output. Do NOT use for tweet/X post writing — use felo-twitter-writer instead. Explicit commands: /felo-superagent.

2026-04-02252

gemini-summary

duotify/GitHubClawToolkit

Use this skill when a user provides a web page URL, PDF file, video, or audio file and wants a Traditional Chinese summary. Prefer this skill for requests like "summarize this URL", "summarize this PDF", "summarize this video", "summarize this audio", "幫我摘要這篇", "這份報告重點是什麼", "幫我總結這段影片", "幫我摘要這段錄音", or when analyzing articles, documents, presentations, YouTube videos, audio recordings, or any content that needs a zh-TW summary. Automatically detects input type (web page, PDF, video, audio) — no manual type selection needed.

2026-04-01252

felo-search

duotify/GitHubClawToolkit

Felo AI real-time web search for questions requiring current/live information. Triggers on current events, news, trends, real-time data, information queries, location queries, how-to guides, shopping, or when Claude's knowledge may be outdated.

2026-03-31252

name	gemini-nanobanana
description	Use this skill when users ask to generate, edit, or compose images with Nano Banana (Gemini image models), including text-to-image, image editing, multi-image composition, grounding, and output sizing/saving controls.

Do this first

Use a Node.js wrapper (@google/genai) as the primary flow (multi-turn edits, grounding, advanced generationConfig).
Use scripts/gemini-nanobanana-cli.js for quick single-turn generation/editing runs.
Do not teach raw curl; keep guidance in JS CLI/wrapper form.

Enforce these defaults

API key source: NANOBANANA_GEMINI_API_KEY with fallback GEMINI_API_KEY.
Default model: gemini-3.1-flash-image-preview (allow env override via NANOBANANA_MODEL).
Reference images: support up to 14 total.
Thinking strength: configurable (minimal|low|medium|high), default High.
Aspect ratio: default Auto.
Resolution: default 1K.
Output mode: default Images only.
Google Search grounding tool: default Disabled; enable with --google-search when needed.
Output directory: default nanobanana-output/, unless the prompt explicitly asks for another location.
512px rule: send imageConfig.imageSize as string "512" in API calls (never numeric 512).

Read references intentionally

Start in references\image-generation-api.md for operational payload rules, model behavior, sizing, thinking, grounding, and limits.
Use references\sources.md to verify source provenance and jump to upstream docs.
For canonical API behavior, read:

Output formatting rules

After successful generation, always report results in this format:
1. A short completion message (e.g. ✅ 圖片已產出)
2. The relative path to each generated file, formatted as a markdown image embed: ![description](relative/path/to/image.jpg)
3. File metadata: format (JPEG/PNG), dimensions, file size

Example final answer:

✅ 圖片已產出

![一杯抹茶拿鐵](workspaces/issue-2/nanobanana-output/matcha-latte-01.jpg)

- 格式：JPEG · 1408×768 · 757 KB

This ensures downstream systems (GitHub Issue comments, Telegram relay) can display or link to the image.

Execution pattern

Default to Node.js wrapper flows for regular usage, especially when payload control is needed.
Quick path (agent runs from workspaces/issue-N/; resolve the repo root first):
- REPO_ROOT=$(git rev-parse --show-toplevel 2>/dev/null || echo "../..") then node "$REPO_ROOT/.agents/skills/gemini-nanobanana/scripts/gemini-nanobanana-cli.js" --prompt "..."
- Add references via repeated -i/--image (up to 14).
- Enable grounding via --google-search when prompt needs fresh web context.
The API key (GEMINI_API_KEY / NANOBANANA_GEMINI_API_KEY) is injected by the workflow environment; do not hardcode it. The CLI reads it automatically from the environment.