← Back to Skills Marketplace
zhuchenggong19851114-design

Auto Contents

by zhuchenggong19851114-design · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
76
Downloads
0
Stars
1
Active Installs
1
Versions
Install in OpenClaw
/install auto-contents
Description
MakeContents AI 资讯 Agent 技能。当用户需要自动执行资讯拉取、筛选推送(AINews/AITopics/AITools)、内容创作、内容发布、内容归档等任务时使用此技能。此技能调用本地运行的 MakeContents 应用 API,无需人工干预即可完成完整的内容创作和分发流程。
README (SKILL.md)

MakeContents Agent 技能

概述

MakeContents 是一个本地运行的 RSS 资讯聚合与内容创作应用。此技能让 Agent 能够自主完成以下工作:

  1. 拉取资讯 — 从配置的 RSS/RSSHub 信源获取最新内容
  2. 筛选推送 — 根据学习到的规律选择有价值的资讯,直接生成推送内容并分发到微信和飞书知识库
  3. 内容创作 — 为选定资讯生成小红书风格的图文内容,自动渲染封面和详情图
  4. 内容归档 — 将创作结果保存到飞书多维表,通知人类审核

配置

在开始前确认以下配置(通过应用系统配置页面或 .env 设置):

  • MakeContents 服务地址:默认 http://localhost:3710
  • 各功能所需的 API Key(飞书、微信)已在应用配置页填写
  • 飞书多维表 URL 已配置(用于内容归档,可选)
  • 飞书机器人 Webhook 已配置(用于完成通知)
  • 小红书 Cookie 已配置 + 发布开关已开启(用于发布笔记,默认关闭,需人类主动开启)

重要提示

  • 必须使用 exec + curl 调用 API,不要使用 web_fetch(后者用于抓取外部网页,无法访问本地服务)
  • 所有 API 请求需包含 Header:Content-Type: application/jsonAccept: application/json
  • 执行前建议先检查服务健康:curl -s http://localhost:3710/api/health 应返回 {"status":"ok"}

API 参考

完整 API 文档见 references/api.md,包含所有接口的请求/响应格式。

选择规律记忆

Agent 的学习规律存储在技能文件路径下 references/agent-rules.md,每次学习后更新此文件。


执行流程

流程一:学习选择规律(可选)

先读取文件references/agent-rules.md文档,如果包含内容筛选规律则使用此规律作为后续任务的选择标准,跳过此步骤;无可参考的规律内容时按顺序执行:

  1. 调用 GET /api/news/agent-summary 获取学习数据
  2. 分析数据:
    • saved_news:用户推送过的资讯(含 push_type)
    • all_news_in_period:对应时段所有资讯
    • 对比两个列表,提炼被选中资讯的共同特征
  3. 将分析的规律追加到 references/agent-rules.md

分析角度

  • 被选中资讯的来源分布(哪些信源更受偏好)
  • 被选中资讯的主题特征(关键词、话题类型)
  • 推送类型分布(用户倾向于哪类推送)
  • 内容创作选题特征(哪类话题被创作)

流程二:资讯推送

To execute the news push workflow:

  1. 调用 POST /api/news/fetch 拉取最新资讯
  2. 调用 GET /api/news/grouped?agent=1 获取资讯列表(已推送条目自动排除,无需手动去重)
  3. 使用 references/agent-rules.md 中的筛选规律
  4. 根据规律从列表中筛选 0-5 条有价值的资讯(无有价值资讯可直接结束此流程)
  5. 对每条资讯:
    • 判断推送类型(ainews / aitopics / aitools)
    • 生成 news_title(≤30字,精炼有力)和 news_summary(100-200字,符合对应风格)
    • 调用对应推送接口(注意:接口路径为 /api/news/{id}/ainews 等,不是 /{id}/ainews/):
      POST /api/news/{id}/ainews
      POST /api/news/{id}/aitopics
      POST /api/news/{id}/aitools
      
    • 请求体示例:{"news_title":"标题","news_summary":"摘要内容"}
  6. 完成后调用 POST /api/content/notify-bot 通知人类

推送类型判断原则

  • ainews:客观事实类资讯,速报风格
  • aitopics:有争议或引人思考的话题,末尾加引导性问题
  • aitools:新工具/新产品发布,突出功能和使用场景

流程三:内容创作

To execute the content creation workflow:

  1. 从资讯列表选定 0-3 条适合内容创作的资讯(通常是有深度或有趣的话题,无有价值资讯可直接结束此流程)
  2. 对每条资讯,构造以下参数并调用 POST /api/content/agent-render
    • news_id:资讯 ID
    • source_url:原始链接(description中的链接,不是资讯的链接;有则填,服务会自动截图)
    • cover_word:1-2个英文单词,概括核心主题
    • cover_title:≤15字,有冲击力的中文标题
    • cover_description:≤20字,补充说明
    • cover_emoji:1个相关 Emoji
    • content_typenews/tools/topics/default(决定配色)
    • title:≤20字,小红书风格标题
    • content:正文,500-800字,小红书风格,有观点有情感
    • tags:3-5个标签,逗号分隔
  3. 获取渲染结果的 saved_content_idcover_urldetail_urls
  4. (可选,未明确则跳过)调用 POST /api/content/save-to-bitable 存入飞书多维表(传 saved_content_id
  5. 调用 POST /api/content/publish-xhs将笔记内容发布到小红书:
    {
      "title": "{title(≤20字)}",
      "desc": "{content}",
      "cover_url": "{cover_url}",
      "detail_urls": ["{detail_urls[0]}", "..."],
      "is_private": true
    }
    
    • Agent 必须传 "is_private": true,笔记将以「仅自己可见」发布,由人类在小红书 App 中审核后手动设为公开
    • 成功后从响应中取 note_url
    • 若接口返回错误"小红书发布功能未开启",不重试、不报错,跳过此步骤即可
  6. 调用 POST /api/content/notify-bot 发送通知,消息格式:
    ✅ 内容创作完成
    
    标题:{title}
    类型:{content_type}
    标签:{tags}
    
    已存入飞书多维表,请审核。
    {如已发布到小红书:🔗 小红书笔记(仅自己可见,请在 App 审核后设为公开):{note_url}}
    

注意事项

  • 截图详情图可能因目标网站防爬而失败,失败时静默跳过,只生成封面
  • 推送接口会自动标记资讯为已保存,无需手动调用 save 接口
  • Agent 生成内容时,news_titlenews_summary 直接作为推送内容,需保证质量
  • 每次执行完整流程后,建议执行一次学习流程以更新规律记忆
  • 小红书发布:发布开关默认关闭,由人类决定是否授权。Agent 应尊重此开关,收到"未开启"错误时静默跳过,不得反复重试或提示人类开启
  • 服务不可用:调用 API 前建议先 GET /api/health。若请求返回 5xx 错误,可能是服务临时故障,可等待数秒后重试一次;若仍失败则终止流程并通知人类
Usage Guidance
This skill is internally consistent: it expects a MakeContents server running on http://localhost:3710 and uses that service (which should hold any API keys/cookies). Before enabling: (1) ensure you actually run the MakeContents app locally and that its stored API keys/cookies are configured securely; (2) be aware the skill asks the agent to run shell commands (exec + curl) — grant that only if you trust the agent runtime; (3) the agent will append to references/agent-rules.md (persistent file writes) — back up this file if you want history control; (4) the Xiaohongshu publish flow requires a separate app-level switch and cookies — the agent will not bypass that switch per its instructions. If any of these behaviors are unacceptable, do not install or restrict the agent's ability to execute shell commands and write files.
Capability Analysis
Type: OpenClaw Skill Name: auto-contents Version: 1.0.0 The skill bundle instructs the AI agent to use high-risk execution methods, specifically 'exec' combined with 'curl', to interact with a local API service (http://localhost:3710). While these capabilities are aligned with the stated purpose of automating RSS news aggregation and social media publishing (WeChat, Feishu, Xiaohongshu), granting an agent shell-level execution access is a significant security risk. There is no evidence of intentional malice, such as data exfiltration or backdoors, but the architectural pattern in SKILL.md and references/api.md relies on a broad attack surface that could be exploited via prompt injection.
Capability Assessment
Purpose & Capability
Name/description say the agent will drive a local MakeContents app and the SKILL.md only calls http://localhost:3710 endpoints and reads/writes the included references files — these are coherent and proportional to the stated purpose.
Instruction Scope
Instructions are specific and limited to the local MakeContents API and the skill's own references files. Two items to note: (1) the SKILL.md explicitly requires using shell exec + curl instead of web_fetch, which means the agent will run shell commands (a broader capability than an in-process fetch); (2) the agent is instructed to append/update references/agent-rules.md (persistent file writes). Both are consistent with the skill's function but are privileges the user should be comfortable granting.
Install Mechanism
Instruction-only skill with no install spec and no third-party downloads — lowest-risk install posture.
Credentials
The skill does not declare or require any environment variables or external credentials. It expects the local MakeContents app to already hold service API keys/cookies in its own configuration, which is appropriate: the agent calls that local service and does not need the secrets itself.
Persistence & Privilege
always is false and the skill does not request system-wide changes. It does instruct the agent to update its own references/agent-rules.md (local persistent memory), which is expected for a learning workflow and limited in scope.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install auto-contents
  3. After installation, invoke the skill by name or use /auto-contents
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
makecontents skill v1.0.0 – 自动化资讯拉取、筛选、创作与分发全流程 - 新增自动资讯获取、筛选、内容创作与归档功能,无需人工干预 - 支持对 RSS/RSSHub 资讯的推送与内容分发到微信、飞书、小红书等平台 - 引入基于用户习惯的智能筛选与推送规律记忆机制 - 提供 API 调用、服务健康检查和详细内容创作流程说明 - 增加飞书多维表归档与小红书笔记私有发布支持 - 加强错误处理和消息通知,支持流程中断与人工审核
Metadata
Slug auto-contents
Version 1.0.0
License MIT-0
All-time Installs 1
Active Installs 1
Total Versions 1
Frequently Asked Questions

What is Auto Contents?

MakeContents AI 资讯 Agent 技能。当用户需要自动执行资讯拉取、筛选推送(AINews/AITopics/AITools)、内容创作、内容发布、内容归档等任务时使用此技能。此技能调用本地运行的 MakeContents 应用 API,无需人工干预即可完成完整的内容创作和分发流程。 It is an AI Agent Skill for Claude Code / OpenClaw, with 76 downloads so far.

How do I install Auto Contents?

Run "/install auto-contents" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Auto Contents free?

Yes, Auto Contents is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Auto Contents support?

Auto Contents is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Auto Contents?

It is built and maintained by zhuchenggong19851114-design (@zhuchenggong19851114-design); the current version is v1.0.0.

💬 Comments