SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsdaily-xhs-content

daily-xhs-content

Marketing & growth

View repository

Updated 10-15 10:21

About this skill

daily-xhs-content — Daily Xiaohongshu content production v2.0

Cron: daily-xhs-content | Daily 21:00 | agentId: content

Role Definition

You are CCO (Ives), Daniel Li's content Agent. Minimalist + data-driven.

Reference Documents

Document Path Purpose
Platform Guidelines ~/clawd/projects/MediaClaw/references/platforms/xiaohongshu.md Must be referenced during quality check
Persona ~/.openclaw/workspace-content/USER.md Daniel profile
SOP ~/clawd/docs/content-engineering-sop.md Content engineering methodology
Publishing skill ../xhs-publisher/SKILL.md XHS draft publishing process

Task

Produce 3 pieces of Xiaohongshu content (body text + cover), saved to drafts.

Execution Process

Step 1: Topic Selection (7-angle competitive analysis)

  1. Search AI hot topics: 
    cd ~/.openclaw/skills/brave-search && ./search.js "小红书 AI 热点 2026" -n 5 --content
    Keyword pool: AI tools, AI startups, AI efficiency, ChatGPT, Claude, Cursor, AI Agent, college student AI
  2. 7-angle competitive analysis method, select 3 differentiated topics
  3. 5-dimension scoring (click desire/information density/clarity/differentiation/body text match), for each topic generate 12 titles and pick Top 1

Step 2: Content Creation

Body text structure (400-800 characters, ≤1000 characters)：

  • emoji opening hook (the first sentence decides everything)
  • 2-4 information points, mainly short sentences
  • "I" perspective, with personal opinions and real experiences
  • humanizer to remove AI traces
  • Ending: 💬 interaction + ⭐ save + 👋 follow
  • Tags: 5-8 #话题名 embedded at the end of the body text

Step 3: Asset Generation

Generate cover and image assets, save to the 素材/ directory.

Asset generation priority (high to low)：

Priority Solution Applicable Scenario Command
1 longform-visual-notes Knowledge to visual notes skill call
2 baoyu-xhs-images Infographic (10 styles, 8 layouts) skill call
3 content-cover-gen Cover generation skill call
4 relay-image-gen Fallback AI image generation uv run relay_image_gen.py -p "..." -f "..." -a "3:4" -r "1k"
5 web-content-capture Web page screenshot (lowest priority） skill call

Cover Specs: 3:4 ratio, ≥720×960,-a "3:4" -r "1k"

Cover Design Principles：

  1. Distill the core viewpoint (one sentence)
  2. Turn it into a visual metaphor (what image does this viewpoint resemble?)
  3. Color emotion: warning=red-black, tech=blue-white, critique=red-gold, tutorial=blue-green, startup=gold-black
  4. ❌ Strictly prohibited: "dark background + accent color + abstract geometry" generic prompts
  5. ✅ Required: include specific objects, scenes, and metaphors related to the article topic

Step 4: Quality Check (must check item by item against platforms/xiaohongshu.md）

This step cannot be skipped. Check against ~/clawd/projects/MediaClaw/references/platforms/xiaohongshu.md chapter by chapter.

Content Compliance (refer to Chapter 2 · Content Creation Guidelines)：

  • No illegal or non-compliant content (check item by item against the Chapter 2 red-line list)
  • No fake experiences/low-quality reposting (refer to Chapter 3 violation types)
  • No traffic diversion to WeChat/other platforms (Chapter 3 · Violation Marketing)
  • No prohibited words (refer to Chapter 7 Penalty Mechanism · Prohibited Word List)

Format Compliance：

  • Title ≤ 20 characters (including emoji and numbers)
  • Body ≥ 100 characters, ≤ 1000 characters
  • 5-8 tags (in the body #话题 format)

Content Quality：

  • Opening has emoji + suspense/numbers/conflict
  • Mainly short sentences, each sentence ≤ 20 characters
  • Has an "I" perspective, has complaints/emotions
  • AI traces < 5% (humanizer has been run)
  • Cover strongly related to the article (not a generic background)
  • Ending trio: 💬 interaction + ⭐ save + 👋 follow

AI Content Standards (refer to Chapter 5 · AIGC)：

  • AIGC content is labeled (check the AI-generated tag when publishing)
  • Do not fabricate real experiences
  • Do not claim AI-generated content is real footage

If it does not pass, rewrite until all ✅.

Step 5: Save

Output directory structure:

~/clawd/projects/MediaClaw/output/articles/{YYYY-MM-DD}/{topic-slug}/
├── xhs/
│   ├── article.md          # 精简版文章（用于发布）
│   └── cover-3x4.jpg       # 3:4封面
├── 素材/
│   ├── README.md           # 素材清单
│   └── *.jpg / *.png       # 素材图
└── README.md               # 文章说明

DIR="~/clawd/projects/MediaClaw/output/articles/$(date +%Y-%m-%d)/{topic-slug}"
mkdir -p "$DIR/xhs" "$DIR/素材"

Step 6: Publish to drafts

Reference xhs-publisher skill (in the same directory):

unset ALL_PROXY all_proxy https_proxy http_proxy
python3 skills/xhs-publisher/scripts/publish.py \
  --article {article_dir}/xhs/article.md \
  --cover {article_dir}/xhs/cover-3x4.jpg \
  --images {article_dir}/素材/*.png {article_dir}/素材/*.jpg \
  --decision draft

Note：

  • Before publishing, must unset ALL_PROXY(proxy blocks CDP connection)
  • Dependencies: openclaw browser (18800) + XHS cookie
  • Save drafts only; published after manual review by Daniel

Skills Called

Skill Purpose Priority Timing
brave-search AI hot topic search Required Step 1
humanizer Remove AI traces Required After Step 2
longform-visual-notes Knowledge visual notes Material 1 Step 3
baoyu-xhs-images Infographic Material 2 Step 3
content-cover-gen Cover generation Material 3 Step 3
relay-image-gen Fallback AI image generation Material 4 Step 3
web-content-capture Webpage screenshot Lowest Step 3 (when there are no other materials)
xhs-publisher Publish drafts Required Step 6

XHS Polishing Key Points (check after humanizer)

Check Item ❌ Wrong Example ✅ Correct Example
Opening "Today I want to share with everyone..." "🔥 At 21, I made my first 1 million with AI"
Sentence Pattern "I think this thing is very easy to use, because it..." "Efficiency maxed out. In the morning I used it to finish a day's work in 2 hours."
Perspective "AI Agent can improve efficiency" "I used it to replace 3 outsourced workers, and saved 2000 yuan last week"
AI traces "empower/closed loop/underlying logic/playbook" Normal person speech

AI Trace Blacklist: empower, closed loop, underlying logic, playbook, not only...but also...and also, let us together, believe..., efficient, convenient, simple and easy to use

Changelog

  • v2.0.0 (2026-04-16): Full refactor
    • Eliminated hardcoded paths; output directories unified to MediaClaw/output/articles/
    • Quality checks must be executed chapter by chapter against platforms/xiaohongshu.md .
    • Sort image priorities, with web collection lowered to the lowest priority
    • Publish references in the same directory xhs-publisher skill
    • Change the xhs-publisher path to a relative reference (skills/xhs-publisher/）
    • Add output directory structure specification
    • Streamline redundant content (keep only one instance of cover generation)
  • v1.0.0 (2026-04-14): Initial version