Token导航 LogoToken导航

Agent Skill · 研究检索

content-engine(内容引擎)

用于辅助文档、README、Markdown、说明文和内容稿件的整理与改写。它适合让 Agent 提炼结构、补齐章节、统一术语、检查链接或把零散材料整理成可读文档。使用时应保留项目已有事实、命令和路径,不要把未确认的信息写成确定结论;涉及对外文案时,还需要控制语气,避免过度营销或夸大能力。

content-engine 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为只读;许可证为 MIT-0,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 5,583。

总安装 25,239周安装 1,021GitHub Stars 公开资料未说明下载 7,923核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT-0
版本
1.0.0

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:content-engine(内容引擎)
来源仓库:https://github.com/ariktulcha/content-engine
安装命令:
openclaw skills install content-engine
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install content-engine

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

全栈内容引擎从竞品分析到 SEO 优化完成端到端内容生产闭环。识别内容差距并生成关键词布局建议,提升搜索引擎可见性。支持多格式输出与 A/B 测试素材准备,加速内容投放决策周期。安装命令:openclaw skills install content-engine,需配置爬虫与索引服务权限。

使用前应确认是否允许整理竞争对手网站公开数据用于策略制定。

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 OpenClaw;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 openclaw skills install content-engine;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

content-engine 怎么安装?

在终端执行 openclaw skills install content-engine。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

content-engine 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

content-engine 需要什么权限?

权限级别是只读。该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

content-engine 是免费的吗?

来源标注的许可证是 MIT-0。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

content-engine 用的人多吗?

公开数据是累计安装 25,239 次,近一周安装 1,021 次,下载 7,923 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

content-engine 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。
  • 标注为只读权限,试用阶段不容易误改本地文件。

先想清楚的地方

  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 OpenClaw,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
content-engine

内容引擎 · 本页

25,239—只读MIT-0通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
content-engine
description
>
metadata
openclaw
emoji
✍️

Content Engine

From blank page to published, optimized, and promoted — in one workflow. This skill turns a topic or keyword into a researched, drafted, optimized, and publish-ready piece of content.

Why This Exists

Content creation with OpenClaw today requires manually chaining 4-5 skills: web research, writing, SEO optimization, CMS formatting, and social scheduling. This skill connects the full pipeline so you go from idea to published post in one flow.

The Pipeline

Content Engine runs in 5 phases. The user can run the full pipeline or start from any phase.

Phase 1: Research

When the user provides a topic or target keyword:

  1. Competitor analysis: Use web_search to find the top 5-10 ranking articles for the target keyword
  2. Structure extraction: For each competitor article, note:

- Word count (approximate from snippets) - H2/H3 headings and structure - Key angles and arguments - What's missing or weak

  1. People Also Ask: Search for "[keyword]" and extract related questions
  2. Content gap identification: What do all competitors cover? What does nobody cover? The gap is the opportunity.
  3. Research brief output:
📊 Research Brief: [Keyword]

Top competitors (by ranking):
1. [Title] — [URL] — ~[word count] words
   Key angle: [one sentence]
2. ...

Common structure:
- All cover: [topics everyone mentions]
- Gap opportunities: [topics nobody covers well]

People Also Ask:
- [question 1]
- [question 2]
- [question 3]

Recommended angle: [your unique take based on gaps]
Recommended word count: [based on competitor average + 20%]

Phase 2: Draft

Generate a structured first draft using the research brief:

  1. Check for brand voice: look in OpenClaw memory for stored brand guidelines, tone preferences, or writing style notes. If none exist, ask the user on first run and store for future use.
  2. Outline first: generate an outline with H2/H3 structure before writing. Show the user and get approval (or auto-proceed if they said "just write it").
  3. Write the draft following these principles:

- Open with a hook that addresses the reader's problem directly - Use the gap opportunities from research as unique sections - Include data points and specific examples (from research) - Write for the target keyword naturally — no keyword stuffing - End with a clear conclusion and call-to-action

  1. Output: Markdown file saved to workspace

Phase 3: Optimize

SEO and readability optimization:

  1. Meta description: Generate a compelling meta description under 155 characters that includes the target keyword
  2. Title tag: Optimize the title for search (include keyword, keep under 60 chars, make it compelling)
  3. Internal link suggestions: if the user has provided a sitemap or list of existing content, suggest internal links. Otherwise, note where internal links could go.
  4. Image alt text: suggest alt text for any images mentioned or planned
  5. Readability check:

- Flag paragraphs longer than 4 sentences - Flag sentences longer than 25 words - Suggest subheadings every 300 words if missing - Check for passive voice overuse

  1. Keyword integration check: verify the target keyword appears in title, first paragraph, at least one H2, and meta description

Output an optimization report appended to the draft:

🔍 SEO Optimization Report

Title tag: [optimized title] ([char count])
Meta description: [meta] ([char count])
Target keyword: [keyword]
  └─ In title: ✅
  └─ In first paragraph: ✅  
  └─ In H2: ✅
  └─ In meta: ✅
Readability: [score/assessment]
Suggested internal links: [list or "provide sitemap for suggestions"]

Phase 4: Format & Publish

Format the content for the user's CMS and prepare for publication:

  1. Detect CMS: check memory for CMS preference. Common options:

- WordPress: use WordPress skill if available, or output HTML-ready content with featured image suggestions - Ghost: output in Ghost-compatible Markdown - Notion: create a Notion page via Notion skill if available - Markdown/Hugo/Jekyll: output as .md with proper frontmatter - No CMS: just output clean Markdown

  1. Frontmatter generation (for static site generators):
   ---
   title: "[optimized title]"
   description: "[meta description]"
   date: [today]
   tags: [relevant tags]
   categories: [relevant categories]
   ---
  1. Publish or save: if CMS integration is available, offer to publish directly. Otherwise, save the final file and tell the user where it is.

Phase 5: Promote

Generate social media promotion content:

  1. Platform-specific posts: generate posts optimized for each platform:

- LinkedIn: professional tone, 1-3 paragraphs, relevant hashtags - Twitter/X: hook + link, under 280 chars, 2-3 hashtags - Reddit: genuine value-add framing (not promotional), suggest appropriate subreddits - Hacker News: technical angle, factual title

  1. Schedule: if Mixpost or Buffer skill is available, offer to schedule posts
  2. Email newsletter: offer to generate a newsletter blurb for the article

Output all promotional content in a single block:

📢 Promotion Kit for: [Article Title]

LinkedIn:
[post text]

Twitter/X:
[tweet text]

Reddit (suggested subreddits: r/[sub1], r/[sub2]):
[post text]

Newsletter blurb:
[2-3 sentence summary for email]

Usage Modes

Full Pipeline

User: "Write a blog post about AI agent security best practices" → Run all 5 phases sequentially, showing output at each stage

Research Only

User: "Research what's ranking for 'openclaw tutorial'" → Run Phase 1 only, output the research brief

Draft from Research

User: "I already researched this topic, here are my notes: [notes]. Write the draft." → Skip Phase 1, run Phases 2-5

Optimize Existing Content

User: "Optimize this blog post for SEO" + [attached content] → Skip Phases 1-2, run Phases 3-5

Promote Existing Content

User: "Generate social posts for this article: [URL or content]" → Skip Phases 1-4, run Phase 5 only

Content Calendar

If the user asks for a content plan or calendar:

  1. Research trending topics in their niche using web_search
  2. Cross-reference with their existing content (if known) to avoid duplication
  3. Suggest 4-8 topics for the next month with:

- Target keyword - Estimated search volume (use web research clues) - Difficulty assessment (how strong is the competition?) - Recommended publish date

  1. Store the calendar in memory for tracking

Storing Brand Context

On first use, ask the user about their brand voice and store in memory:

  • Tone: professional, casual, technical, friendly, authoritative?
  • Audience: developers, marketers, business owners, general public?
  • Formatting preferences: short paragraphs? lots of headers? code examples?
  • Things to avoid: jargon level, competitors not to mention, topics to skip
  • Existing content URL: for internal linking and avoiding duplication

Once stored, use these preferences for every future content generation without asking again.

Edge Cases

  • No keyword given: if the user just says "write about AI agents", help them choose a specific keyword first using research
  • Very competitive keyword: warn the user and suggest long-tail alternatives
  • Existing content: if the user's site already has a similar article, flag it and suggest updating instead of creating new
  • Multiple languages: support content creation in any language the user requests, adjusting SEO practices for that language's search engine norms
  • Short-form content: for social posts or email copy (not blog posts), skip Phases 1 and 3, go straight to writing + formatting

平台分布

OpenClaw

74.64%
按下载量换算5,914

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

未展示

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills