功能描述

使用火山引擎豆包模型生成图片。通过火山引擎豆包图片生成 API 创建图片。支持自定义提示词、尺寸、模型等参数。使用方式：生图：一只可爱的小猫。

使用说明 (SKILL.md)

doubao-image - 豆包图片生成

Name: doubao-image
Author: wanxing0724

使用火山引擎豆包模型生成图片。

简介

通过火山引擎豆包图片生成 API 创建图片。支持多模型自动切换、URL 输出等功能。

环境变量

变量名	必填	说明
`VOLCENGINE_IMAGE_API_KEY`	Yes	豆包 API Key
`DOUBAO_IMAGE_OUTPUT_DIR`	No	图片输出目录（默认：workspace/downloads/images）

Get API Key:

Login https://console.volces.com/
Go to API Key management
Copy your API Key

使用方式

生图：一只可爱的小猫

命令行参数

参数	说明	默认值
prompt	图片描述	(必填)
--model	使用的模型	doubao-seedream-4-0-250828
--size	图片尺寸	1024x1024
--num	生成数量	1
--url-only	仅返回 URL，不下载	false

支持的模型（自动切换）

按优先级排序，优先使用第一个，失败则自动尝试下一个：

doubao-seedream-4-0-250828 (默认)
doubao-seedream-4-5-251128
doubao-seedream-5-0-260128
doubao-seedream-3-0-t2i-250415

支持的尺寸

1024x1024 (默认)
1280x720
720x1280
1024x768
768x1024

输出格式

控制台输出

==================================================
Prompt: a cute kitten
Size: 1024x1024
Num: 1
==================================================
Trying model: doubao-seedream-4-0-250828
[OK] Model doubao-seedream-4-0-250828 success

==================================================
Success!
Model: doubao-seedream-4-0-250828
URL: https://...
==================================================

Downloading to: C:\Users\zcf\.openclaw\workspace\downloads\images\doubao_1234567890.jpeg
[OK] Saved: C:\Users\zcf\.openclaw\workspace\downloads\images\doubao_1234567890.jpeg
  Size: 256.5 KB

[FILE_PATH] C:\Users\zcf\.openclaw\workspace\downloads\images\doubao_1234567890.jpeg

特殊标记

[URL_ONLY] \x3Curl> - 仅生成 URL 时的输出
[FILE_PATH] \x3Cpath> - 文件保存路径

输出目录

图片自动保存到 downloads/images/ 目录。

可通过环境变量自定义：

# Windows
set DOUBAO_IMAGE_OUTPUT_DIR=D:\my_images

# Linux/Mac
export DOUBAO_IMAGE_OUTPUT_DIR=/path/to/images

重要说明

API Key: 从 VOLCENGINE_IMAGE_API_KEY 环境变量读取
Multi-model fallback: 如果默认模型失败，会自动尝试备选模型
URL Output: 生成后立即输出图片 URL

错误处理

如果所有模型都失败，会显示具体错误信息
如果下载失败，仍会输出 URL 供手动下载

安全使用建议

This skill appears to implement exactly what it claims (calls VolcEngine's Doubao image API and optionally saves images), but the package metadata does not declare the required VOLCENGINE_IMAGE_API_KEY env var and the script contains a hardcoded Windows default output path (C:/Users/zcf/.openclaw/...). Before installing: 1) Confirm you are comfortable providing your VolcEngine API key (it will be sent as a Bearer token to ark.cn-beijing.volces.com). Prefer creating a key with minimal scope. 2) If you don't want files written to a default user path, set DOUBAO_IMAGE_OUTPUT_DIR or OPENCLAW_WORKSPACE to a safe directory. 3) Consider running the script in a restricted/sandboxed environment or reviewing the code yourself (scripts/generate.py is short and readable). 4) Ask the publisher/registry to correct the manifest to declare the required env var and document OPENCLAW_WORKSPACE usage so metadata accurately reflects runtime requirements.

功能分析

Type: OpenClaw Skill Name: doubao-image-openclaw Version: 1.0.4 The skill is a legitimate tool for generating images via the Volcengine Doubao API. The Python script `scripts/generate.py` correctly implements API calls to the official endpoint and handles image downloads as described. While it contains a hardcoded default path referencing a specific local user directory ('zcf') in the fallback logic, this is a minor portability issue rather than a security risk or malicious behavior.

能力评估

ℹ Purpose & Capability

The code (scripts/generate.py) and SKILL.md both implement image generation against VolcEngine's Doubao image API and list the same models/sizes; this matches the declared purpose. However, the registry metadata claims no required environment variables while both SKILL.md and the script require VOLCENGINE_IMAGE_API_KEY, an inconsistency between claimed requirements and actual needs.

✓ Instruction Scope

Runtime instructions and the script stay within the stated purpose: they post prompts to the VolcEngine image API, handle responses, and optionally download produced images. The script does not attempt to read unrelated system files or contact unexpected external endpoints beyond the VolcEngine API and image URLs returned by it.

✓ Install Mechanism

This is instruction-only with a single Python script and no install spec, so nothing is downloaded or installed automatically by the skill manifest. That lowers installation risk compared to packages that fetch archives or run installers.

⚠ Credentials

The script and SKILL.md require a secret-like env var VOLCENGINE_IMAGE_API_KEY (and optionally DOUBAO_IMAGE_OUTPUT_DIR). The manifest metadata lists no required env vars/primary credential — this omission is problematic because the skill will fail or will attempt to use your API key if provided. The script also reads OPENCLAW_WORKSPACE (not documented in registry metadata). No unrelated credentials are requested, but the metadata mismatch is a proportionality and transparency concern.

✓ Persistence & Privilege

The skill does not request 'always: true' or any elevated persistence. It writes generated images to a local output directory (default or via DOUBAO_IMAGE_OUTPUT_DIR/OPENCLAW_WORKSPACE) but does not modify other skills or system-wide configuration.

版本历史

v1.0.4

Version 1.0.3 changelog: - Added multi-model automatic fallback: if the default model fails, alternative models are tried sequentially. - Added support for outputting only the image URL without downloading. - Allowed custom output image directory via `DOUBAO_IMAGE_OUTPUT_DIR` environment variable. - Improved output with special tags for URL and file path. - Documentation updated for clearer usage steps, command-line arguments, and output format details.

v1.0.3

Version 1.0.3 changelog: - Added multi-model automatic fallback: if the default model fails, alternative models are tried sequentially. - Added support for outputting only the image URL without downloading. - Allowed custom output image directory via `DOUBAO_IMAGE_OUTPUT_DIR` environment variable. - Improved output with special tags for URL and file path. - Documentation updated for clearer usage steps, command-line arguments, and output format details.

v1.0.2

修改了描述

元数据

Slug doubao-image-openclaw

版本 1.0.4

许可证 MIT-0

累计安装 1

当前安装数 1

历史版本数 3

常见问题

doubao-image 是什么？

使用火山引擎豆包模型生成图片。通过火山引擎豆包图片生成 API 创建图片。支持自定义提示词、尺寸、模型等参数。使用方式：生图：一只可爱的小猫。它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 544 次。

如何安装 doubao-image？

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

doubao-image 是免费的吗？

是的，doubao-image 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。

doubao-image 支持哪些平台？

doubao-image 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 doubao-image？

由 wanxing0724（@wanxing0724）开发并维护，当前版本 v1.0.4。

doubao-image