← 返回 Skills 市场
144
总下载
1
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install nano-banana-2-pro
功能描述
Generate/edit images with Nano Banana 2 (Gemini 3.1 Flash Image). Use for image create/modify requests incl. edits. Supports text-to-image + image-to-image;...
安全使用建议
This skill appears to implement exactly what it claims (a Gemini-based image generator), but the package metadata is incomplete and inconsistent: it fails to declare that you must supply a GEMINI_API_KEY (or pass --api-key) and that the 'uv' runner plus Python deps (google-genai, pillow) are required. Before installing or running: 1) Confirm the skill's origin and trustworthiness (homepage/source unknown). 2) Only provide an API key with the correct, least-privilege scope for image generation; avoid reusing high-privilege account keys. 3) Ensure you run it in a safe directory (it will read any input image paths you pass and will create output files). 4) Consider installing dependencies from official sources and running the script in a sandbox if origin is untrusted. 5) Ask the publisher to update the registry metadata to declare required binaries and the GEMINI_API_KEY requirement and to fix the inconsistent path/slug to reduce ambiguity.
功能分析
Type: OpenClaw Skill
Name: nano-banana-2-pro
Version: 1.0.0
The skill is a legitimate tool for generating and editing images using Google's Gemini API. The core logic in `scripts/generate_image.py` uses the official `google-genai` SDK and `pillow` library to process requests, with no evidence of data exfiltration, malicious execution, or prompt injection. It handles API keys and file outputs using standard practices for CLI utilities.
能力评估
Purpose & Capability
The skill's stated purpose (image generation/editing via Gemini) matches the code and instructions: the script uses google.genai to call Gemini and Pillow to process images. However the registry metadata does not declare the actual runtime needs: the SKILL.md and script require the 'uv' runner and a Gemini API key, but required binaries/env vars are listed as none. Also SKILL.md uses a path (~/.codex/skills/nano-banana-pro/...) that does not exactly match the skill slug (nano-banana-2-pro), which could indicate sloppy packaging or a copy-paste error.
Instruction Scope
The SKILL.md and script limit activity to reading input images from user-specified paths, calling the Google GenAI client with provided API key, and writing output PNGs to the user's working directory. There are no instructions to read unrelated system files or to send data to unexpected external endpoints beyond the Gemini API. The script prints load/save messages and model text parts but does not attempt hidden exfiltration. Still, the instructions grant the script the ability to read any input image paths the user provides, which is expected for this purpose.
Install Mechanism
There is no install spec in the registry (instruction-only), but the script header and runtime import show real dependencies (google-genai, pillow) and the SKILL.md presumes the 'uv' command. The skill does not declare how those dependencies are installed or verified. Lack of an install mechanism plus commented dependency hints is disproportionate: either declare required binaries/env and provide an install, or keep it pure-instructions and make requirements explicit.
Credentials
The script expects an API key (CLI --api-key or GEMINI_API_KEY environment variable) and will fail without it, yet the registry lists no required env vars or primary credential. Requesting a service API key is reasonable for this functionality, but omitting that from metadata is an incoherence that affects user consent and credential management. Additionally, SKILL.md's pre-check mentions 'command -v uv' which means the 'uv' binary is required but not declared.
Persistence & Privilege
The skill does not request persistent/automatic inclusion (always: false) and does not modify other skills or global agent settings. It only writes output files where the user chooses. There are no elevated privileges or persistence indicators.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install nano-banana-2-pro - 安装完成后，直接呼叫该 Skill 的名称或使用
/nano-banana-2-pro触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.0.0
Initial release of Nano Banana 2 skill (Gemini 3.1 Flash Image Preview):
- Added SKILL_CN.md and scripts/generate_image.py.
- Supports text-to-image and image editing using up to 14 input images.
- Allows control over resolution (512/1K/2K/4K), 14 aspect ratios, and multiple models.
- Includes --thinking-level and --image-only options.
- API key management and robust preflight checks.
- Filename generation with timestamp and context-based naming.
元数据
常见问题
nano banana 2 是什么？
Generate/edit images with Nano Banana 2 (Gemini 3.1 Flash Image). Use for image create/modify requests incl. edits. Supports text-to-image + image-to-image;... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 144 次。
如何安装 nano banana 2？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install nano-banana-2-pro」即可一键安装，无需额外配置。
nano banana 2 是免费的吗？
是的，nano banana 2 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
nano banana 2 支持哪些平台？
nano banana 2 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 nano banana 2？
由 gybin02（@gybin02）开发并维护，当前版本 v1.0.0。
推荐 Skills