← Back to Skills Marketplace
68
Downloads
0
Stars
1
Active Installs
1
Versions
Install in OpenClaw
/install keplerjai-image-gen
Description
通过 ThinkZone AI 生成图片。当用户说「画图」「生图」「生成图片」「AI 绘图」「做一张图」「图生图」等时使用。支持 3 个图片模型（Gemini/MiniMax/Seedream）。
README (SKILL.md)
\r \r
ThinkZone Image Generation\r
\r 使用 ThinkZone AI 平台的 3 个指定模型生成图片。\r \r
⚠️ 调用流程（必读）\r
\r 当用户未指定模型时，必须先反问用户要用哪个模型，再执行生成。可这样询问：\r \r
请问要用哪个模型？\r
- 图片：Gemini（多模态/参考图）、MiniMax Image 01（人物主体参考）、Seedream（轻量 2K/3K）\r \r 用户明确指定模型后，再调用对应脚本。\r \r 当用户已指定模型（如「用 Gemini 画」「MiniMax 生成」）时，直接执行，无需反问。\r \r
🎨 支持的模型（3 个）\r
\r
图片生成模型（3 个）\r
\r
| 模型 | Model ID | 说明 | 尺寸/分辨率 |\r
|------|----------|------|------|\r
| Gemini 3.1 Flash Image Preview | gemini-3.1-flash-image-preview | Google 多模态图像生成 | 0.5K, 1K, 2K, 4K |\r
| MiniMax Image 01 | image-01 | MiniMax 图片生成 | 自定义 512-2048px |\r
| BytePlus Seedream 5.0 Lite | doubao-seedream-5-0-260128 | BytePlus 轻量版 | 2K, 3K |\r
\r
🔧 配置\r
\r 需要设置环境变量：\r
THINKZONE_API_KEY- API 密钥\rTHINKZONE_BASE_URL- API 基础 URL（可选，默认https://open.thinkzoneai.com）\r \r
📸 图片生成使用\r
\r
Gemini 3.1 Flash Image Preview\r
\r
# 基本用法\r
python3 {baseDir}/scripts/gen_image.py \\r
--prompt "一只穿宇航服的虾" \\r
--model "gemini-3.1-flash-image-preview"\r
\r
# 指定分辨率和宽高比\r
python3 {baseDir}/scripts/gen_image.py \\r
--prompt "科幻城市夜景" \\r
--model "gemini-3.1-flash-image-preview" \\r
--resolution "2K" \\r
--aspect-ratio "16:9"\r
\r
# 带参考图（支持多张，最多 10 张）\r
python3 {baseDir}/scripts/gen_image.py \\r
--prompt "基于参考图生成" \\r
--model "gemini-3.1-flash-image-preview" \\r
--images "path/to/image1.jpg" "path/to/image2.jpg" \\r
--resolution "1K"\r
```\r
\r
**参数说明：**\r
\r
| 参数 | 说明 | 默认值 | 可选值 |\r
|------|------|--------|--------|\r
| `--prompt` | 图像描述文本 | 必填 | - |\r
| `--model` | 模型名称 | `gemini-3.1-flash-image-preview` | 固定 |\r
| `--resolution` | 分辨率 | `1K` | `0.5K`, `1K`, `2K`, `4K` |\r
| `--aspect-ratio` | 宽高比 | `1:1` | `1:1`, `16:9`, `9:16`, `3:2`, `2:3`, `3:4`, `4:3`, `4:5`, `5:4`, `21:9`, `1:4`, `4:1`, `1:8`, `8:1` |\r
| `--images` | 参考图路径（最多 10 张） | - | 图片路径 |\r
| `--thinking-level` | 思考等级 | `minimal` | `minimal`, `high` |\r
| `--output-dir` | 输出目录 | `./tmp/thinkzone-image` | 路径 |\r
\r
### MiniMax Image 01\r
\r
```bash\r
# 基本用法\r
python3 {baseDir}/scripts/gen_image.py \\r
--prompt "一只可爱的猫咪" \\r
--model "image-01"\r
\r
# 指定宽高和数量\r
python3 {baseDir}/scripts/gen_image.py \\r
--prompt "赛博朋克角色" \\r
--model "image-01" \\r
--width 1024 \\r
--height 1024 \\r
--n 4\r
\r
# 图生图（人物主体参考）\r
python3 {baseDir}/scripts/gen_image.py \\r
--prompt "穿古装的女性角色" \\r
--model "image-01" \\r
--subject-reference "path/to/character.jpg" \\r
--aspect-ratio "3:4" \\r
--watermark\r
```\r
\r
**参数说明：**\r
\r
| 参数 | 说明 | 默认值 | 可选值 |\r
|------|------|--------|--------|\r
| `--prompt` | 图像描述文本 | 必填 | - |\r
| `--model` | 模型名称 | `image-01` | 固定 |\r
| `--width` | 宽度像素 | `1024` | 512-2048（8 的倍数） |\r
| `--height` | 高度像素 | `1024` | 512-2048（8 的倍数） |\r
| `--aspect-ratio` | 宽高比 | `1:1` | `16:9`, `4:3`, `1:1`, `3:4`, `9:16`, `21:9`, `2:3`, `3:2` |\r
| `--n` | 生成数量 | `1` | 1-9 |\r
| `--subject-reference` | 人物主体参考图 | - | 图片路径 |\r
| `--watermark` | 是否添加水印 | `false` | `true`, `false` |\r
\r
### BytePlus Seedream 5.0 Lite\r
\r
```bash\r
# 基本用法\r
python3 {baseDir}/scripts/gen_image.py \\r
--prompt "中国风山水画" \\r
--model "doubao-seedream-5-0-260128"\r
\r
# 指定尺寸和格式\r
python3 {baseDir}/scripts/gen_image.py \\r
--prompt "高质量产品渲染图" \\r
--model "doubao-seedream-5-0-260128" \\r
--size "3K" \\r
--output-format "png"\r
\r
# 批量生成（最多 15 张）+ 参考图（最多 14 张）\r
python3 {baseDir}/scripts/gen_image.py \\r
--prompt "基于参考图变体" \\r
--model "doubao-seedream-5-0-260128" \\r
--images "ref1.jpg" "ref2.jpg" \\r
--size "2K" \\r
--max-images 4 \\r
--no-watermark\r
```\r
\r
**参数说明：**\r
\r
| 参数 | 说明 | 默认值 | 可选值 |\r
|------|------|--------|--------|\r
| `--prompt` | 图像描述文本 | 必填 | - |\r
| `--model` | 模型名称 | `seedream-5-0-260128` | 固定 |\r
| `--size` | 输出尺寸 | `2K` | `2K`, `3K` |\r
| `--output-format` | 输出格式 | `jpeg` | `jpeg`, `png` |\r
| `--max-images` | 最大生成数量 | `1` | 1-15 |\r
| `--images` | 参考图路径（最多 14 张） | - | 图片路径 |\r
| `--watermark` | 是否添加水印 | `false` | `true`, `false` |\r
| `--stream` | 流式输出 | `false` | `true`, `false` |\r
\r
---\r
\r
## 📋 前端调用参考（Vue 3）\r
\r
参考 `campus-amags/tenant/src/views/workbench` 组件的调用方式：\r
\r
### 图片生成调用示例\r
\r
```typescript\r
import { \r
postImageGeneration, // MiniMax Image 01\r
postV3ImageGenerations, // BytePlus Seedream\r
postGeminiGenerateContent // Gemini 3.1 Flash\r
} from './api/image'\r
\r
// MiniMax Image 01\r
const minimaxPayload = {\r
model: 'image-01',\r
prompt: '你的提示词',\r
width: 1024,\r
height: 1024,\r
aspect_ratio: '1:1',\r
n: 1,\r
aigc_watermark: false,\r
subject_reference: [{ type: 'character', image_file: 'base64_or_url' }]\r
}\r
const result = await postImageGeneration(minimaxPayload)\r
\r
// BytePlus Seedream 5.0 Lite\r
const seedreamPayload = {\r
model: 'doubao-seedream-5-0-260128',\r
prompt: '你的提示词',\r
size: '2K',\r
output_format: 'jpeg',\r
watermark: false,\r
image: ['url1', 'url2'], // 参考图\r
sequential_image_generation: 'auto',\r
sequential_image_generation_options: { max_images: 4 }\r
}\r
const result = await postV3ImageGenerations(seedreamPayload)\r
\r
// Gemini 3.1 Flash Image Preview\r
const geminiPayload = {\r
contents: [\r
{ parts: [{ text: '你的提示词' }] },\r
{ parts: selectedImages.map(img => ({ inlineData: { mimeType: 'image/jpeg', data: img.base64 } })) }\r
],\r
generationConfig: {\r
responseModalities: ['IMAGE'],\r
imageConfig: {\r
aspectRatio: '16:9',\r
imageSize: '2K'\r
},\r
thinkingConfig: {\r
thinkingLevel: 'minimal',\r
includeThoughts: false\r
}\r
}\r
}\r
const result = await postGeminiGenerateContent(geminiPayload)\r
```\r
\r
---\r
\r
## ⚠️ 注意事项\r
\r
1. **账户余额**：确保 ThinkZone AI 账户有足够余额\r
2. **图片尺寸限制**：\r
- Gemini：分辨率 0.5K/1K/2K/4K\r
- MiniMax：512-2048px（8 的倍数）\r
- Seedream：2K/3K\r
3. **参考图限制**：\r
- Gemini：最多 10 张\r
- MiniMax：1 张（人物主体）\r
- Seedream：最多 14 张\r
4. **超时处理**：图片生成建议预留约 120 秒\r
\r
---\r
\r
## 🔗 相关链接\r
\r
- ThinkZone AI 平台：https://open.thinkzoneai.com\r
- 前端参考实现：`C:\Users\Linsihuan\Desktop\campus-amags enant\src\views\workbench`\r
\r
---\r
\r
*最后更新：2026-04-17*\r
Usage Guidance
This skill appears to implement ThinkZone image generation, but there are several red flags you should consider before installing or enabling it:
- The repository includes multiple plaintext API keys across files (API_KEY.md, API_KEY_SETUP.md, TEST_SUCCESS.md, scripts/test_models.py, and a hardcoded fallback in scripts/gen_image.py). Do not trust embedded keys — they may be stale, shared, or tied to third-party accounts. Remove any embedded secrets and supply your own API key via environment variables.
- The code intentionally falls back to a hardcoded API key when THINKZONE_API_KEY is not set. That behavior means the skill can use a secret baked into the skill bundle. If you install this skill, insist that it be modified to refuse to run without a valid environment-provided key.
- Documentation and code disagree about supported models and the skill's deployment scope (3 vs 5 models; some docs claim global registration). Ask the publisher to provide a single authoritative SKILL.md and remove conflicting materials.
- ENV_FIX.md recommends machine-level env variables and wrapper scripts. Avoid setting machine-scoped secrets unless you fully trust and control the service/account. Prefer per-service (agent) keys and least-privilege configuration.
- Because the repo contains real-looking API keys, consider rotating any keys used in your environment if you previously tested with this package. Treat those keys as compromised.
What to do before installing:
1) Request a cleaned package with no embedded secrets and a clear SKILL.md that matches the registry metadata.
2) Verify the skill will not use any fallback hardcoded key (remove fallback in gen_image.get_config()).
3) Confirm the endpoint base URL and model IDs are what you expect.
4) If you must test, run in an isolated environment and use a disposable API key.
Given these inconsistencies and embedded secrets, I rate this skill as suspicious and recommend remediation or a vendor clarification before enabling it in production.
Capability Analysis
Type: OpenClaw Skill
Name: keplerjai-image-gen
Version: 1.0.0
The skill bundle contains multiple instances of hardcoded API keys (e.g., 'amags_b6324919...') across several files including scripts/gen_image.py, scripts/test_models.py, and API_KEY.md. While the core functionality of image generation via ThinkZone AI appears legitimate, the explicit inclusion of credentials and specific local file paths (e.g., C:\Users\Linsihuan\...) indicates poor security hygiene and potential credential exposure. No evidence of intentional data exfiltration or malicious execution was found, but the hardcoded secrets pose a significant security risk.
Capability Tags
Capability Assessment
Purpose & Capability
Name/description: image generation via ThinkZone AI (3 models). The repository/docs are inconsistent about supported models (SKILL.md says 3 models; README/VERIFICATION/DONE/test scripts advertise/support 5 models). Several files (README, VERIFICATION, TEST_SUCCESS) claim the skill is globally registered/usable by all agents, but registry metadata flags 'always' as false. These mismatches suggest sloppy packaging and uncertain scope.
Instruction Scope
SKILL.md runtime instructions themselves are generally scoped to asking the user for model choice and calling the scripts. However other docs/instructions (ENV_FIX.md, API_KEY.md, DONE.md) instruct administrators to set system/global environment variables, use wrapper scripts, or even hardcode API keys into code. That expands the operational scope beyond what's declared and could cause credentials to be stored in plaintext on the host.
Install Mechanism
No install spec is provided and the skill is instruction + script only, so nothing is downloaded or installed automatically. This is the lowest-risk install mechanism.
Credentials
The declared required env var is a single API key (THINKZONE_API_KEY), which is reasonable. But the codebase includes multiple hardcoded API keys in plaintext (API_KEY.md, API_KEY_SETUP.md, DONE.md, ENV_FIX.md, TEST_SUCCESS.md, scripts/gen_image.py fallback, scripts/test_models.py). There are several distinct keys across files. Embedding working API keys and recommending hardcoding in the script is disproportionate and risky (exposes secrets in the repo, encourages privilege escalation to set machine-level env vars).
Persistence & Privilege
Registry flags show normal privileges (always:false, agent-invocable). However multiple documentation files assert global availability and global registration, and ENV_FIX.md suggests setting machine-level env vars and creating wrapper scripts — practices that increase the skill's system footprint if followed. The skill does not itself request 'always: true', but the documentation encourages global/system configuration.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install keplerjai-image-gen - After installation, invoke the skill by name or use
/keplerjai-image-gen - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
- Initial release of keplerjai-image-gen, enabling image generation via ThinkZone AI.
- Supports three image models: Gemini 3.1 Flash Image Preview, MiniMax Image 01, and BytePlus Seedream 5.0 Lite.
- Requires users to explicitly choose a model before generating, with model-specific parameter options.
- Detailed usage instructions and parameter tables provided for all models.
- Environment variable THINKZONE_API_KEY required.
- Includes frontend integration references for Vue 3.
Metadata
Frequently Asked Questions
What is Keplerjai Image Gen?
通过 ThinkZone AI 生成图片。当用户说「画图」「生图」「生成图片」「AI 绘图」「做一张图」「图生图」等时使用。支持 3 个图片模型（Gemini/MiniMax/Seedream）。 It is an AI Agent Skill for Claude Code / OpenClaw, with 68 downloads so far.
How do I install Keplerjai Image Gen?
Run "/install keplerjai-image-gen" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Keplerjai Image Gen free?
Yes, Keplerjai Image Gen is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Keplerjai Image Gen support?
Keplerjai Image Gen is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Keplerjai Image Gen?
It is built and maintained by 任佶 (@renjicode); the current version is v1.0.0.
More Skills