AgentSkillsCN

csdn-blog-publisher

为 CSDN 撰写技术博客文章,并可选择发布。单篇文章即完整的工作流程(撰写 + 发布)。系列文章或批量投稿则只需撰写(将 .md 文件本地保存,后续再手动发布)。

SKILL.md
--- frontmatter
name: csdn-blog-publisher
description: Write and optionally publish technical blog posts for CSDN. Single post = full workflow (write + publish). Series/batch posts = write only (save .md locally for manual publishing).

CSDN Blog Writer & Publisher

Generate high-quality technical blog posts in Markdown format for CSDN.

Workflow Modes

Mode 1: Single Post (Full Workflow)

code
1. Research → 2. Write Draft → 3. Open CSDN Editor → 4. Import & Configure → 5. Publish

Use this when user asks to write AND publish ONE specific blog post.

Mode 2: Series/Batch Posts (Write Only)

code
1. Research → 2. Write Draft → 3. Save to Output Directory

Use this when user asks to create multiple posts or a blog series. User handles publishing manually.


Step 1: Research Content

Gather authoritative sources using web_fetch:

python
# Recommended sources by topic
SOURCES = {
    "ai_agent": [
        "https://github.com/WooooDyy/LLM-Agent-Paper-List",
        "https://arxiv.org/abs/2309.07864",  # Agent Survey
        "https://github.com/langchain-ai/langchain",
    ],
    "llm": [
        "https://github.com/Hannibal046/Awesome-LLM",
        "https://arxiv.org/list/cs.CL/recent",
    ],
    "general": [
        "https://github.com/trending",
        "https://news.ycombinator.com",
    ]
}

Fetch and extract key information:

code
web_fetch(url, extractMode="markdown", maxChars=50000)

Step 2: Write Blog Draft

2.1 Blog Structure

Follow the format in references/blog-format.md.

Essential elements:

  • 📑 Table of contents with emoji section headers
  • 💡 "思考" (Thought) questions for reader engagement
  • 📊 Comparison tables (use bullet lists for Discord/WhatsApp compatibility)
  • 💻 Code examples with syntax highlighting
  • 📚 Numbered references [1], [2], [3]...
  • ASCII diagrams for architecture illustrations
  • ⚠️ MUST end with "## 参考文献" (References) section

Target: 10,000-20,000 Chinese characters for comprehensive posts.

2.2 Title Format

For series posts:

code
【{Series}篇】{Number}:{Title}

Example:

code
【Agents篇】04:Agent 的推理能力——思维链与自我反思

2.3 Content Guidelines

  1. Opening: Hook with a relatable scenario or question
  2. Body:
    • Use hierarchical headings (##, ###)
    • Include 💡思考/🤔解答 pairs for engagement
    • Add code examples with explanations
    • Use ASCII art for diagrams (CSDN renders these well)
  3. Closing:
    • Summary + future outlook
    • MUST include "## 参考文献" as the final section

Mode 1 Only: CSDN Publishing Steps

Skip this section for batch/series posts

Step 3: Open CSDN Editor & Import

python
# Navigate to CSDN Markdown editor
browser(action="open", profile="openclaw", targetUrl="https://editor.csdn.net/md?not_checkout=1")

# Import content via file upload or paste

Step 4: Configure & Publish

  1. Set Title (5-100 characters)
  2. Upload Cover Image (from Pixabay or similar free source)
  3. Add Tags (e.g., "人工智能")
  4. Generate Summary (AI extract or manual, 256 chars max)
  5. Click Publish

Step 5: Verify Publication

  • Check URL changed to article page
  • Verify content is complete (not truncated)
  • Note the article URL for records

Mode 2 Only: Save Output

For series/batch posts, save to the specified output directory:

python
# Default output directory
OUTPUT_DIR = "/home/jianxiong/doc"

# Filename format
filename = f"{Series}_{Number:02d}_{ShortTitle}_Blog.md"

# Example
# Agents_04_Reasoning_Blog.md

User will manually publish these files to CSDN.


Output Checklist

For All Posts:

  • Blog content is 10,000-20,000 characters
  • Has table of contents
  • Has at least 5 major sections
  • Includes code examples
  • Ends with "## 参考文献" section
  • Saved to output directory

For Single Post (Mode 1) Only:

  • Cover image uploaded
  • Tags configured
  • Article published to CSDN
  • Article URL obtained

Series: Agents 专栏

Based on LLM-Agent-Paper-List:

#TitleKey TopicsStatus
01AI Agent 的崛起综述、架构
02Agent 的大脑LLM核心能力
03Agent 的记忆系统Memory, RAG, MemGPT
04Agent 的推理能力CoT, ToT, Self-Refine
05Agent 的规划能力ReAct, Plan-and-Execute
06Agent 的感知模块多模态输入
07Agent 的行动模块Tool Use, Embodied
08单智能体应用代码、研究、游戏
09多智能体协作CAMEL, ChatDev
10人机协作Human-in-the-Loop
11LangChain & LangGraph框架深度解析
12AutoGen多智能体对话
13MetaGPT软件公司模式
14CrewAI角色团队协作
15实战:构建Agent从零搭建
16Agent 训练RLHF, AgentGym
17Agent 评估Benchmarks
18Agent 社会行为、人格模拟
19具身智能机器人控制
20Agent 的未来挑战与展望

Resources