← 返回 Skills 市场
434
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install dragon-xiaohongshu
功能描述
自动发布内容到小红书平台。支持发布图文、检查登录状态、获取登录二维码。使用场景:自动化小红书内容发布、批量发布、定时发布等。
使用说明 (SKILL.md)
小红书自动发布 Skill
自动化发布内容到小红书平台,支持图文发布、登录管理和批量操作。
功能特性
- ✅ 自动发布图文内容到小红书
- ✅ 检查登录状态
- ✅ 自动获取登录二维码
- ✅ 支持自定义标题、内容、图片和标签
- ✅ 支持 HTTP MCP 服务端模式
前置要求
-
小红书 MCP 服务端
- 文件:
xiaohongshu-mcp-windows-amd64.exe - 端口:18060
- 启动方式:运行 exe 文件
- 文件:
-
mcporter CLI
npm install -g mcporter -
已登录小红书账号
- 首次使用需要先扫码登录
快速开始
1. 启动 MCP 服务端
# 在 xiaohongshu-mcp 目录下运行
.\xiaohongshu-mcp-windows-amd64.exe
2. 检查登录状态
node scripts/publish.js --check
3. 发布内容
# 使用默认配置
node scripts/publish.js
# 自定义内容
node scripts/publish.js \
--title "我的标题" \
--content "正文内容..." \
--image "C:\\path\ o\\image.png" \
--tags "tag1,tag2,tag3"
配置说明
mcporter 配置
配置文件:~/.mcporter/mcporter.json
{
"mcpServers": {
"xiaohongshu": {
"url": "http://localhost:18060/mcp"
}
}
}
发布参数
| 参数 | 类型 | 必填 | 说明 |
|---|---|---|---|
| title | string | 是 | 标题(最多20字) |
| content | string | 是 | 正文内容(最多1000字) |
| images | array | 是 | 图片路径数组 |
| tags | array | 否 | 标签数组 |
使用示例
示例 1:发布产品推广
const publish = require('./scripts/publish');
await publish({
title: '新品推荐!超好用',
content: '今天给大家推荐一款超好用的产品...',
images: ['C:\\product.jpg'],
tags: ['好物推荐', '新品', '种草']
});
示例 2:发布生活分享
node scripts/publish.js \
--title "今天的心情日记" \
--content "今天天气真好,出去散步..." \
--image "C:\\photos\ oday.jpg" \
--tags "生活,日常,心情"
注意事项
- 登录状态:发布前必须已登录小红书
- 图片格式:支持 jpg、png 格式
- 内容限制:
- 标题最多 20 个中文字
- 正文最多 1000 字
- 图片最多 9 张
- 频率限制:小红书有每日发帖限制(约50篇)
故障排除
问题:MCP 连接失败
解决:
- 检查 MCP 服务端是否运行
- 检查端口 18060 是否被占用
- 重启 MCP 服务端
问题:未登录
解决:
- 运行
node scripts/publish.js --check检查状态 - 使用登录工具扫码登录
- 或运行
node scripts/login.js获取二维码
问题:发布失败
解决:
- 检查图片路径是否正确
- 检查内容是否超过字数限制
- 查看 MCP 服务端日志
参考资料
- 小红书 MCP 项目
- mcporter 文档
- 详细配置见
references/config.md
安全使用建议
Before installing or running this skill: 1) Understand it depends on a separate MCP server binary (xiaohongshu-mcp-windows-amd64.exe) and the mcporter CLI — these are not installed by the skill and are not declared in the registry metadata. Only obtain the MCP executable from a trusted source (the SKILL.md links to a GitHub repo; review that repo and its releases carefully). 2) Note that SKILL.md references node scripts/login.js to get a QR code, but login.js is not included — expect missing functionality or manual login steps. 3) The scripts operate on local image files and talk only to localhost:18060; they do not request external credentials, but you should verify the MCP server’s behavior and permissions before running an untrusted .exe (it could perform network or privileged actions). 4) Remove or inspect hard-coded absolute paths in example files before running (they reference the original developer's workspace). 5) If you need this skill, prefer to manually install and run the MCP server after auditing its code/binaries; if you cannot verify the MCP binary/source, do not run it. If you want a cleaner integration, ask the skill author to: declare required binaries/env in metadata, include a login.js if referenced, and avoid hard-coded user-specific paths.
功能分析
Type: OpenClaw Skill
Name: dragon-xiaohongshu
Version: 1.0.0
The skill requires the user to manually download and execute an external, pre-compiled Windows executable (`xiaohongshu-mcp-windows-amd64.exe`) as a prerequisite, as explicitly stated in `SKILL.md` and `references/config.md`. The core script (`scripts/publish.js`) then communicates with this executable via `http://localhost:18060/mcp`. While the `publish.js` script itself does not exhibit malicious behavior (e.g., data exfiltration, persistence, unauthorized network calls beyond localhost), the reliance on an unverified external binary introduces a significant trust and security risk, as the security posture of the entire skill is dependent on the integrity and behavior of this third-party executable.
能力评估
Purpose & Capability
The skill's name/description (automated Xiaohongshu publishing) aligns with the code and instructions: the scripts post to a local MCP service on localhost:18060 and provide publish/check features. However, SKILL.md requires a Windows binary (xiaohongshu-mcp-windows-amd64.exe) and the mcporter CLI, but the registry metadata declares no required binaries/env — this mismatch is unexpected and should have been declared in metadata.
Instruction Scope
Runtime instructions and scripts only interact with local resources: they read image files, call a localhost HTTP endpoint (/mcp), and output console logs. There is no obvious network exfiltration to remote hosts beyond localhost. Two issues: SKILL.md suggests running node scripts/login.js to get a QR code, but no login.js is present in the package; and several example posts reference absolute local paths under the author's workspace, which may not exist for an installer. These are scope/integrity problems (missing file and hard-coded paths), not direct evidence of exfiltration.
Install Mechanism
This is instruction-only from the skill registry (no install spec), so nothing is automatically downloaded by the skill bundle. However, the instructions require the user to obtain/run a third‑party Windows executable (xiaohongshu-mcp-windows-amd64.exe) and to install mcporter globally (npm). The skill points to a GitHub repo for 'xiaohongshu-mcp' which is the likely source — the user must manually vet and obtain that executable. Relying on an external unsigned .exe introduces risk if the binary's provenance is not confirmed.
Credentials
The registry lists no required environment variables or credentials, and the code communicates with a local MCP endpoint rather than remote APIs that would need service keys. References/config.md mentions an MCP_URL env var as optional. No secret/credential requests are present in code or metadata. This is proportionate, but the metadata omission (not declaring required tools/binaries) reduces transparency.
Persistence & Privilege
The skill does not request permanent presence (always: false) and does not modify other skills or system-wide settings. It is user-invocable and can be run by the agent; autonomous invocation is allowed by default but is not combined with other high-risk flags here.
如何使用
- 确保已安装 OpenClaw(本地或 Docker 部署)
- 在对话框中输入安装命令:
/install dragon-xiaohongshu - 安装完成后,直接呼叫该 Skill 的名称或使用
/dragon-xiaohongshu触发 - 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
初始版本
元数据
常见问题
小红书发布 是什么?
自动发布内容到小红书平台。支持发布图文、检查登录状态、获取登录二维码。使用场景:自动化小红书内容发布、批量发布、定时发布等。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 434 次。
如何安装 小红书发布?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install dragon-xiaohongshu」即可一键安装,无需额外配置。
小红书发布 是免费的吗?
是的,小红书发布 完全免费(开源免费),可自由下载、安装和使用。
小红书发布 支持哪些平台?
小红书发布 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。
谁开发了 小红书发布?
由 dragon015(@dragon015)开发并维护,当前版本 v1.0.0。
推荐 Skills