← Back to Skills Marketplace
xiaohongshu-cover-gen
by
velmavalienteqejimu22-jpg
· GitHub ↗
· v1.0.0
· MIT-0
101
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install xiaohongshu-cover-gen
Description
当用户需要为小红书帖子生成封面图和内容图卡时,加载此 Skill。触发词:小红书配图、封面生成、图卡设计、XHS cover、social media image。
Usage Guidance
This skill plausibly does what it claims (research -> prompt -> Lovart image generation -> PIL text cards), but there are important red flags to consider before installing or running it:
- Credentials: The skill requires Lovart authentication cookies (a JWT-like 'usertoken' plus two other cookies) and instructs you to save them in a local file (.lovart_cookies.json). The registry metadata did not declare any required credential — treat that as a mismatch. Do not provide your Lovart cookies to untrusted code or public/shared agents.
- Sensitive file handling: The skill reads/writes the cookies file and downloads images to disk. If you must use it, keep the cookies file in a secure, access-restricted location and delete/rotate the token after use.
- Insecure TLS workaround: Scripts set NODE_TLS_REJECT_UNAUTHORIZED=0 and use rejectUnauthorized: false to fetch images. That disables certificate validation and makes downloads vulnerable to man-in-the-middle attacks. Prefer an environment where TLS works correctly, or ask the author for an alternative (official API or properly signed endpoints).
- Metadata mismatch: The skill's metadata claims no required binaries/credentials while the README and SKILL.md list agent-browser, Node.js, Python+Pillow, and Lovart cookies. Ask the publisher to correct the metadata and explicitly declare required credentials.
- Code review / sandboxing: There is executable code (shell, Node, Python). Review the scripts before running them. For safety, run the skill in an isolated environment/container and avoid giving it persistent platform privileges or access to unrelated credentials.
- Small correctness note: the included Python generator contains an apparent string/quote bug in the quote-drawing section that should be fixed before running.
If you trust the author and accept the above trade-offs, use short-lived cookies, run in a sandbox, and rotate tokens after each session. If you do not trust the source, do not provide your Lovart auth tokens or run the scripts with network/file access.
Capability Analysis
Type: OpenClaw Skill
Name: xiaohongshu-cover-gen
Version: 1.0.0
The skill bundle facilitates automated image generation but employs several high-risk security practices. Specifically, `scripts/download_image.sh` explicitly disables SSL certificate validation (NODE_TLS_REJECT_UNAUTHORIZED=0), which is a significant vulnerability. Furthermore, `references/lovart-operation.md` instructs the agent to extract and store sensitive session cookies (usertoken, useruuid) in a local file to maintain authentication. While these behaviors are documented as functional requirements for interacting with the Lovart platform, the combination of credential handling and weakened transport security poses a risk to the user's session integrity.
Capability Assessment
Purpose & Capability
Name/description match the included workflow, docs, and helper scripts (Python PIL, agent-browser, Node.js). However the registry metadata claims no required binaries or credentials while the SKILL.md and README explicitly require agent-browser, Node.js, Python+Pillow, and a Lovart account + exported cookies. That mismatch (metadata says 'none' but docs require multiple tools and credentials) is inconsistent.
Instruction Scope
SKILL.md guides the agent to obtain and store Lovart authentication cookies (usertoken JWT, useruuid, webid) in a local .lovart_cookies.json file, inject cookies via agent-browser or document.cookie, fetch/download images via browser eval and Node, and write files to disk (/tmp, output/). Those actions go beyond mere prompt-writing: they require handling user auth tokens and performing network/file I/O. Collecting/storing JWT cookies is sensitive and is not declared in the skill metadata.
Install Mechanism
There is no install spec (instruction-only plus helper scripts). That keeps install risk low. The included scripts are plain shell/Node/Python files rather than downloads from untrusted URLs.
Credentials
The skill requires Lovart authentication (cookies/JWT) at runtime, but the declared requirements list no env vars or primary credential. The code and docs instruct users to create a local .lovart_cookies.json with a long JWT and to inject it. Asking for authentication tokens is expected for automating a third‑party web UI, but the skill should have declared this as a required credential and explained safe storage/rotation. Additionally the scripts disable TLS verification (NODE_TLS_REJECT_UNAUTHORIZED=0) to download images — a high-risk concession that weakens transport security.
Persistence & Privilege
always:false and no modifications to system-wide or other skills' configurations. The skill does instruct writing a cookies JSON file and output images to local directories (normal for this workflow) but does not request persistent platform privileges.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install xiaohongshu-cover-gen - After installation, invoke the skill by name or use
/xiaohongshu-cover-gen - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
AI Agent Skill:为小红书帖子生成高质量封面图和内容图卡
经 5 轮实战验证、6 次翻车迭代沉淀的配图方法论。 不是"调 API 生成图片"——是一套从审美调研、创意构思、prompt 编写到生成交付的完整工作流。
✨ 特性
🎯 6 阶段标准流程:调研 → 分析 → 确认 → prompt → 生成 → 检查
🧠 14 条实战翻车经验(Gotchas):按严重程度分级,避免重复踩坑
🖌️ 核心审美原则:从用户情绪出发设计封面,而非技术概念
🤖 Lovart 完整操作手册:agent-browser 自动化 Phase 0-5 全流程
✅ 8 项 prompt 自检 checklist:发送前最后一道质量关卡
📊 审美积累体系:定期浏览 + 分类归档,持续提升审美水平
🐍 Python 图卡生成器:PIL 批量生成纯文字卡,Design Token 可配置
📁 目录结构
xiaohongshu-cover-gen/
├── SKILL.md # ⭐ 核心 Skill 文件(加载此文件)
├── README.md # 项目说明
├── .gitignore
│
├── references/ # 参考文档
│ ├── lovart-operation.md # Lovart 平台完整操作手册
│ ├── prompt-checklist.md # Prompt 发送前 8 项自检
│ ├── research-platforms.md # 调研平台操作指南
│ └── aesthetic-guide.md # 审美积累文档模板 & 指南
│
├── scripts/ # 辅助脚本
│ ├── gen_text_cards.py # Python 图卡生成器
│ ├── check_jwt.sh # Lovart JWT 过期检查
│ └── download_image.sh # Lovart 图片下载(绕过 SSL)
│
└── templates/ # 模板文件
├── design-tokens.md # Design Token 模板 + 3 种配色示例
└── iteration-log.md # 迭代日志模板
🚀 Quick Start
1. 作为 Agent Skill 使用
将 SKILL.md 加载到你的 AI Agent(如 CodeBuddy、Cursor、Claude 等),Agent 会自动按照 Skill 中定义的流程工作。
触发词:小红书配图、封面生成、图卡设计、XHS cover、social media image
2. 首次使用前准备
Lovart 账号(生成封面图)
注册 Lovart 账号
在浏览器 DevTools 中导出 3 个关键 cookies:usertoken、useruuid、webid
保存为 .lovart_cookies.json(格式见 references/lovart-operation.md)
Python 环境(生成文字图卡)
pip install Pillow
3. 工作流概览
收到帖子内容
↓
阶段 0:精准调研(10min)— Dribbble + 小红书 + 品牌官网
↓
阶段 1:帖子分析 → 配图方案
↓
阶段 2:方案确认 + 素材需求前置
↓
阶段 3:写 prompt(过 8 项自检)
↓
阶段 4:Lovart 生成 / Python 图卡
↓
阶段 5:整合检查
🛠️ 前置依赖
依赖 用途 必需?
Lovart 账号 AI 生图 ✅ 封面图必需
agent-browser 浏览器自动化 ✅ Lovart 操作必需
Node.js JWT 检查 + 图片下载 ✅
Python 3 + Pillow 文字图卡生成 可选(纯文字卡时用)
📖 核心理念
封面只有 0.5 秒的生存时间
小红书信息流中,用户拇指飞速滑动。封面的职责不是"完整传达信息",而是在 0.5 秒内制造好奇心和点击欲。
从用户情绪出发,不是从技术概念出发
这是 5 轮迭代最核心的认知沉淀。
❌ 错误:帖子讲 AI 论文 → 画一个大脑(最直线的联想 = 最缺乏创意的路径)
✅ 正确:帖子讲"马斯克点赞的 AI 论文" → 画一个 3D Clay 马斯克竖大拇指
审美 > 工具
配图能力的核心不是工具操作,而是审美和生态理解。工具可以学会,审美才是壁垒。
⚠️ 已知限制
Lovart 不支持 headless 浏览器 OAuth 登录,必须手动提供 cookies
Lovart JWT 有效期约 7 天,需定期更新
agent-browser 命令长度限制 ≤ 1024 字节,长 prompt 需分段输入
agent-browser wait 超过 30s 会触发 idle timeout,需分段等待
Lovart 连续操作可能触发 hCaptcha,建议每次会话只生 1-2 张
a.lovart.ai SSL 证书与 curl/Python 不兼容,图片下载必须用 Node.js
📝 License
MIT
🙏 致谢
Lovart — AI 设计生成工具
Dribbble — 设计灵感来源
Pillow (PIL) — Python 图像处理
5 轮实战中踩过的所有坑 🕳️
Metadata
Frequently Asked Questions
What is xiaohongshu-cover-gen?
当用户需要为小红书帖子生成封面图和内容图卡时,加载此 Skill。触发词:小红书配图、封面生成、图卡设计、XHS cover、social media image。 It is an AI Agent Skill for Claude Code / OpenClaw, with 101 downloads so far.
How do I install xiaohongshu-cover-gen?
Run "/install xiaohongshu-cover-gen" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is xiaohongshu-cover-gen free?
Yes, xiaohongshu-cover-gen is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does xiaohongshu-cover-gen support?
xiaohongshu-cover-gen is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created xiaohongshu-cover-gen?
It is built and maintained by velmavalienteqejimu22-jpg (@velmavalienteqejimu22-jpg); the current version is v1.0.0.
More Skills