XHS Auto-Publish

该 Skill 使用 playwright-core 连接已启用 CDP 且已登录小红书创作服务平台的浏览器，自动上传 1-18 张图片并填写标题、正文和话题标签，支持先截图预览或直接发布，侧重小红书图文笔记的社媒内容发布，不包含账号登录与内容生成能力。

jerryops2026-del

热度 77.05

最近更新 2026-05-18 12:08:08

分类：营销依赖：联网浏览搜索 / 代码执行 / 读写文件应用场景：社交媒体 / 自动化脚本 / 文案创作

安装技能

复制下面的话术发给你的 Agent，它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill： npx secure-skills-cli add xhs-auto-publish

Metadata

name: xhs-auto-publish
description: Automate publishing image+text posts (图文笔记) on Xiaohongshu (小红书) creator platform. Use when the user wants to publish a Xiaohongshu post, create XHS content, auto-post to 小红书, or schedule social media posts for the Chinese market. Supports multi-image upload, title, body text, and hashtags. Requires a browser with CDP (Chrome DevTools Protocol) already logged in to creator.xiaohongshu.com.

小红书自动发布 (Xiaohongshu Auto-Publish)

Publish image+text posts to Xiaohongshu via browser automation. No API key needed — uses CDP to control an already-logged-in browser.

Prerequisites

playwright-core installed: npm install -g playwright-core
A Chromium browser running with CDP (remote debugging) enabled
Logged in to creator.xiaohongshu.com in that browser

For OpenClaw users: the managed browser (profile="openclaw") provides CDP automatically at http://127.0.0.1:18800.

Quick Start

node scripts/publish.js \
  --title "你的标题（20字以内）" \
  --body "正文内容（1000字以内）" \
  --images cover.png,slide2.png,slide3.png \
  --hashtags "AI绘画,独立开发" \
  --cdp-url http://127.0.0.1:18800

This auto-publishes the post. Add --dry-run to preview without publishing.

Options

Flag	Description	Required
`--title <text>`	Post title, max 20 Chinese chars	✅
`--body <text>`	Post body, max 1000 chars	✅*
`--body-file <path>`	Read body from a text file	✅*
`--images <paths>`	Comma-separated image paths (1-18)	✅
`--hashtags <tags>`	Comma-separated hashtags (auto-prefixed with #)	Optional
`--cdp-url <url>`	CDP endpoint (default: `http://127.0.0.1:9222`, or `XHS_CDP_URL` env)	Optional
`--dry-run`	Preview only, skip publish (default: auto-publish)	Optional
`--screenshot <path>`	Save preview screenshot path	Optional

*One of --body or --body-file is required (unless using --hashtags alone).

Workflow

Connects to browser via CDP
Navigates to creator.xiaohongshu.com/publish/publish
Clicks "上传图文" (upload images) tab
Uploads all images via file input
Sets title (native input value setter for React compatibility)
Sets body (innerHTML on TipTap/ProseMirror editor)
Takes preview screenshot
Optionally clicks "发布" (publish) button

Tips

Title: Keep under 20 chars. Hook-style titles work best (e.g. "用AI画塔罗牌也太好看了吧")
Body: Short paragraphs with emoji. 小红书 is scroll-heavy — hooks > long text
Images: First image is the cover. Use 3:4 ratio (1242×1660) for best display
Hashtags: Add 3-5 relevant tags for discoverability
Preview first: Always run without --publish first to verify via screenshot

Troubleshooting

Problem	Solution
"Cannot connect to CDP"	Ensure browser is running with remote debugging. For OpenClaw: check `openclaw browser status`
"Redirected" / not logged in	Open the browser manually and log in to creator.xiaohongshu.com
Title input not found	Page may not have loaded. Increase wait time or check network
Images fail to upload	Ensure images are PNG/JPG/WebP and under 20MB each

XHS Content Best Practices

Short hook titles > descriptive titles (20 char limit forces creativity)
3-6 images per post is the sweet spot
First image = cover — make it eye-catching
Short punchy paragraphs with emoji — not walls of text
End with a question to drive comments
3-5 hashtags for discovery, more feels spammy