← 返回 Skills 市场
flywhale-666

Nano Banana Image T8

作者 flywhale · GitHub ↗ · v0.1.8 · MIT-0
cross-platform ⚠ suspicious
260
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install nano-banana-image-t8-win
功能描述
使用提供的 Nano Banana API Key,根据提示词生成或编辑图像,支持文生图和图生图,默认模型为香蕉2,可切换香蕉pro。
安全使用建议
This skill will contact an external image API and persist your Nano Banana API key and a default-model setting under ~/.whaleclaw. Before installing: 1) confirm the source/author (Homepage is missing), 2) verify the API base URL (SKILL.md forces ai.t8star.cn and forbids other domains), 3) be comfortable with the skill saving keys to ~/.whaleclaw/credentials (it sets file perms to owner-only but still stores the secret), and 4) review the included script yourself for any unexpected endpoints or behavior (the code maps user-facing names to internal model IDs). If you can't verify the origin or don't want credentials stored locally, do not install. If you want this cleared to 'benign', provide authoritative publisher metadata or a trusted upstream release URL and update the registry to declare the required env var/credential.
功能分析
Type: OpenClaw Skill Name: nano-banana-image-t8-win Version: 0.1.8 The skill bundle is a legitimate integration for the 'Nano Banana' image generation service (hosted at ai.t8star.cn). The Python script `test_nano_banana_2.py` follows security best practices by implementing cross-platform file permission restrictions (chmod 600 on Linux/macOS and icacls on Windows) for stored API keys. The `SKILL.md` instructions specifically include privacy-preserving constraints that forbid the AI agent from capturing or saving API keys unless the user is explicitly interacting with this specific skill. There is no evidence of data exfiltration, obfuscation, or unauthorized execution.
能力评估
Purpose & Capability
The skill's stated purpose (image generation/editing using a Nano Banana API key) matches the included Python script and SKILL.md instructions. However the registry metadata lists no required environment variables/credentials while the skill clearly requires an API key (param_guard and script). That's an inconsistency between metadata and actual requirements that reduces transparency.
Instruction Scope
Runtime instructions require extracting or accepting an API key from the user and explicitly saving it to ~/.whaleclaw/credentials/nano_banana_api_key.txt, persistently storing a default model, and running the shipped script (which performs network requests). The SKILL.md also imposes detailed capture rules (only extract 'sk-' keys under certain message conditions) and forbids scanning other project files, but the param_guard configuration and the script's save behavior create edges where keys may be captured/saved. The skill asks to limit checks to env var NANO_BANANA_API_KEY and the saved file, but the registry claims no required env; this mismatch is notable.
Install Mechanism
There is no install spec (instruction-only + bundled script). That is low-risk from an install mechanism perspective because nothing is downloaded or executed beyond the included script.
Credentials
The skill requires an API key and persists it to disk with owner-only permissions; this is reasonable for a remote image API. However the registry metadata omitted the required env var/credential. The script will write files under the user's home (~/.whaleclaw), persist the default model, and may overwrite an existing saved key when the user supplies a new one—these are sensitive operations and should be declared upfront in metadata.
Persistence & Privilege
The skill writes persistent configuration to the user's home (~/.whaleclaw/credentials and workspace directories) and may create output images there. It does not request 'always: true' or modify other skills. Persisting an API key and default model is functional but increases the blast radius if the skill or host is compromised—users should be aware the key is stored locally.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install nano-banana-image-t8-win
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /nano-banana-image-t8-win 触发
  4. 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
v0.1.8
- Added comprehensive skill usage instructions for Nano Banana image generation integration (文生图/图生图) with clear parameter requirements. - Enforced strict parameter handling, API Key management, and user interaction rules to ensure only required prompts are used. - Defined clear guidelines for default model handling, switching between “香蕉2” and “香蕉pro,” and model persistence. - Standardized execution to use specific scripts and environment variables, prohibiting ad hoc scripting or direct API calls. - Outlined fixed directory usage for outputs and specific conditions under which to prompt for, save, or reuse API Keys. - Clarified concise, immediate interaction style and explicit error/parameter handling.
元数据
Slug nano-banana-image-t8-win
版本 0.1.8
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 1
常见问题

Nano Banana Image T8 是什么?

使用提供的 Nano Banana API Key,根据提示词生成或编辑图像,支持文生图和图生图,默认模型为香蕉2,可切换香蕉pro。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 260 次。

如何安装 Nano Banana Image T8?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install nano-banana-image-t8-win」即可一键安装,无需额外配置。

Nano Banana Image T8 是免费的吗?

是的,Nano Banana Image T8 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Nano Banana Image T8 支持哪些平台?

Nano Banana Image T8 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 Image T8?

由 flywhale(@flywhale-666)开发并维护,当前版本 v0.1.8。

💬 留言讨论