Long Article Illustration - 长文配图助手

长文配图助手。为长篇文章自动划分段落、生成AI配图提示词并调用图像生成工具完成配图。适用场景：(1) 公众号/博客长文需要配图 (2) 用户上传文章希望获得段落配图建议 (3) 批量生成文章插图 (4) 用户提到"文章配图""长文插画""段落配图"等关键词时触发

Security Scan

The skill's requests and instructions are coherent with its stated purpose (segment long articles, generate image prompts, and call image-generation tools); it is instruction-only, requests no secrets, and poses no unexplained behaviors.

Purpose & Capability

Name and description match the instructions: the skill describes paragraph segmentation, prompt generation, and invoking image-generation tools. There are no unexpected required binaries, env vars, or config paths that don't fit the stated purpose.

Instruction Scope

SKILL.md stays focused on article segmentation and prompt generation and then suggests calling a configured image-generation API, local model, or manual tools. This is in-scope. Note: it instructs the agent to call external image-generation services if configured — those API credentials would be provided outside the skill and are not requested by the skill itself.

Install Mechanism

Instruction-only skill with no install spec and no code files — nothing is downloaded or written to disk by the skill itself, which minimizes install-time risk.

Credentials

The skill declares no required environment variables or credentials. The recommendation to use external image APIs is reasonable for the task; any API keys would be configured separately by the user (not requested by the skill).

Persistence & Privilege

always:false and user-invocable:true (defaults). The skill does not request persistent system-wide changes or special privileges.

Assessment

This skill is coherent and instruction-only: it will ask you to supply the article text and optional style settings, generate per-paragraph prompts, and then either call whatever image-generation API/tool you have configured or output prompts you can copy into a tool. Before installing, consider: (1) privacy — the skill will process the full article text you give it, so avoid submitting sensitive content unless you trust the configured image API; (2) API keys — the skill itself does not request keys, but if you connect DALL·E/Midjourney/Stable Diffusion, keep those credentials secure and scoped; (3) copyright and likeness — generated images may implicate copyrighted characters or real-person likenesses; (4) content policy — ensure your image-generator provider allows the content you intend to create. If you plan to enable autonomous invocation for agents that can call external APIs, be deliberate about which API credentials you provide to the agent.

长文配图助手

为长篇文章自动生成配图，包括智能段落划分、配图提示词生成、图像生成调用。

工作流程

用户上传长文 → 段落划分 → 提示词生成 → 图像生成 → 输出交付

Step 1: 收集输入

向用户确认以下信息：

参数	必填	默认值	说明
文章内容	✅	-	完整文章文本
图片风格	❌	扁平插画	见 references/style-presets.md
图片比例	❌	16:9	公众号内文建议16:9，封面2.35:1
配图密度	❌	每2-3段一图	可选：每段一图/每小节一图

Step 2: 段落划分

按以下优先级划分：

1. 显式标题：识别 #、##、一、、1. 、（1）、（一）、1️⃣、①等标题标记
2. 语义边界：主题转换、时间跳跃、场景切换
3. 长度均衡：单个配图覆盖300-600字为宜

输出格式：

【段落1】标题/摘要
正文内容...

【段落2】标题/摘要
正文内容...

Step 3: 生成配图提示词

为每个需要配图的段落生成提示词，遵循以下结构：

[主体描述], [场景/背景], [风格关键词], [技术参数]

提示词生成规则

1. 提取核心意象：从段落中识别最具画面感的元素
2. 避免抽象概念：将抽象概念转化为具象画面
3. 保持风格一致：全文配图使用统一的风格关键词，特别是出现了人物或角色的图片，需要保持是同一个人物
4. 添加技术参数：比例、光线、构图等
5. 不要出现文字：在提示词中，需要强调最终生成的图片不要包含任何文字

风格预设详见 references/style-presets.md

示例

段落内容：

创业初期，我们挤在一间不到20平米的小办公室里，每天工作到深夜...

生成提示词：

A small crowded office space with desks and computers, warm lamplight at night, papers and coffee cups scattered around, flat illustration style, warm color palette, 16:9

Step 4: 调用图像生成

根据生成的提示词，调用图像生成模型完成配图。

支持的生图方式：

• 调用已配置的 AI 生图 API（如 DALL-E、Midjourney、Stable Diffusion 等）
• 将提示词复制到在线生图工具手动生成
• 使用本地部署的生图模型

💡 用户可根据自身需求选择生图工具，并自行配置对应的 API

Step 5: 输出交付

输出格式选项

选项A：Markdown文档（默认）

# 文章标题

![配图1](image1.png)
第一段正文内容...

![配图2](image2.png)
第二段正文内容...

选项B：提示词清单

| 段落 | 段落摘要 | 配图提示词 |
|------|----------|------------|
| 1 | xxx | prompt1 |
| 2 | xxx | prompt2 |

选项C：打包下载

• 所有图片文件
• 带图片引用的Markdown文档
• 提示词记录文件

参考资料

• 风格预设与提示词模板：references/style-presets.md
• 段落划分详细规则：references/paragraph-rules.md
• 常见问题处理：references/troubleshooting.md