Installable skillMIT

掘金热门文章排行榜

An MCP server that provides on-demand skill loading for AI coding assistants.

BY modbenderUPDATED Oct 28, 2026
INTENDED FOR◈OpenClawInstall in a listed agent.

instructions.md

---
name: juejin-article-trends-cc859059f083
description: 获取掘金网站热门文章排行榜,支持查询文章分类列表和各分类的热门文章趋势。当用户需要了解掘金技术文章排行榜、获取前端/后端/AI等领域的热门文章时使用此技能。
---

# 掘金热门文章排行榜

## 功能概述

此技能用于获取掘金(juejin.cn)网站的技术文章排行榜数据,包括:
- 文章分类列表(前端、后端、AI、移动开发等)
- 各分类的热门文章排行榜
- 文章详细信息(标题、作者、阅读量、点赞数等)

## 工作流程

### 1. 获取分类列表

当用户需要了解掘金文章分类时:

```bash
node scripts/juejin.js categories
```

返回示例:
```json
[
  { "id": "6809637769959178254", "name": "前端" },
  { "id": "6809637769959178255", "name": "后端" },
  { "id": "6809637769959178256", "name": "Android" },
  { "id": "6809637769959178257", "name": "iOS" },
  { "id": "6809637769959178258", "name": "人工智能" },
  { "id": "6809637769959178260", "name": "开发工具" },
  { "id": "6809637769959178261", "name": "代码人生" },
  { "id": "6809637769959178262", "name": "阅读" }
]
```

### 2. 获取热门文章

当用户需要获取特定分类的热门文章时:

```bash
node scripts/juejin.js articles <category_id> [type] [limit]
```

参数:
- `category_id`: 分类ID(从分类列表获取)
- `type`: 排序类型,可选 `hot`(热门) 或 `new`(最新),默认 `hot`
- `limit`: 返回文章数量,默认20

返回示例:
```json
[
  {
    "title": "文章标题",
    "brief": "文章摘要...",
    "author": "作者名",
    "articleId": "123456789",
    "popularity": 100,
    "viewCount": 5000,
    "likeCount": 200,
    "collectCount": 150,
    "commentCount": 50,
    "url": "https://juejin.cn/post/123456789",
    "tags": ["JavaScript", "Vue"]
  }
]
```

## 使用示例

### 查看所有分类
```bash
node scripts/juejin.js categories
```

### 获取前端热门文章(前10篇)
```bash
node scripts/juejin.js articles 6809637769959178254 hot 10
```

### 获取后端最新文章
```bash
node scripts/juejin.js articles 6809637769959178255 new 15
```

## 作者介绍

- 爱海贼的无处不在
- 我的微信公众号:无处不在的技术
COMPATIBILITY & REFERENCES

Platform guidance

Imported from the listed AgentSkill.sh category. Complete supplied package files and notices are retained. Includes source scripts; review them before use. Required external tools and credentials are not installed. No bundled code was executed during import. Individual agent task behavior has not been tested.