xiaohongshu-ops skill
An OpenClaw skill that runs a Xiaohongshu account through the built-in browser: feed analysis, ideas, note drafting, comment replies and a knowledge base
- 2,443 ★ on the repository
- Last commit
Broad access
Why this level
- The skill drives a signed-in account in the browser: it fills the publishing page and can post comments in your name
- The account session is kept in the built-in browser profile, and covers need a Gemini API key
- The agent reads other people's notes and accounts and saves analyses to local files
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/xiaohongshu-opsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The skill drives a signed-in account in the browser: it fills the publishing page and can post comments in your name
- The account session is kept in the built-in browser profile, and covers need a Gemini API key
- The agent reads other people's notes and accounts and saves analyses to local files
This is third-party code. Review the repository files before installing.
What it does
xiaohongshu-ops turns OpenClaw into an assistant for running a Xiaohongshu account. All actions go through the built-in browser with the `openclaw` profile over CDP; the first run requires signing in by scanning a QR code, after which the login is kept. Before any task the agent reads `persona.md` with the account's voice and `knowledge-base/README.md` so as not to repeat what has been analyzed. Analysis has three parts. Home feed analysis shows why high-engagement notes are recommended to you and picks out title hooks and structures. Account analysis takes the 9 to 15 latest notes and assesses positioning, structure, engagement, recognizability and sustainability, and the result holds a main strength, the biggest weakness and a next step. Topic ideation merges platform signals, demand and positioning into 3 to 5 ideas with a hook and a three-part structure. Results are saved in `knowledge-base` as separate files by type: accounts, topics, patterns, actions, reviews. For publishing, the agent prepares the cover, title, text and topics of a note, opens the publishing page and by default stops before the button. The note replication mode, given a link, analyzes the source note and builds a new one with the same idea, by default creating a cover only in a similar style. For comments the agent first checks the notifications page and replies one at a time, verifying the addressee in the input placeholder, with an 8 to 15 second pause between replies. On a platform warning about action frequency it stops.
Who it is for. For creators and marketers who run their own Xiaohongshu account, read Chinese and want to automate analysis, note drafting and comment handling.
Good fit when
- You want to learn from your home feed which title and note structures the platform pushes
- You need to review your own or a public account: what is strong, what is weak and the next step
- You need 3 to 5 ready topics with a hook and structure for the account's positioning
- You want to check new comments and prepare replies in the account's voice with human confirmation
- Analysis results should accumulate in a knowledge base of Markdown files and be reused in later tasks
Not a fit when
- You need other platforms or international social networks: the skill is written only for Xiaohongshu's interface and rules
- You do not work in OpenClaw: all instructions assume its built-in browser and are not verified in other environments
- You need fully unattended posting and mass replies: by default the skill stops before the publish button and answers one comment at a time
- You want to include the skill's code in your own product: the project has no license, so copying it without the author's consent is not allowed
Example request
Analyze my Xiaohongshu home feed and explain why these notes are recommended, then suggest five topics for my accountLimitations
The skill and all its rules, examples and prompts are written in Chinese. It works through the page interface, so when the layout changes the selectors have to be updated: the skill itself says to take a snapshot and refresh the selector rather than push an old scenario. Each action is retried at most once, and after two failures in a row the agent changes route and reports it.
Covers in note replication mode need a Gemini API key (Nano Banana image generation), which the author says can be obtained for free. The latest project updates came in March 2026. The author gave no license, so rights to use and copy are not defined, check with the author. The skill controls an account on a platform, so check the platform's rules on automation and on copying other people's material yourself: the tool builds material modeled on someone else's note and in `tight` mode can give a near copy.
How to disable. Delete the `xiaohongshu-ops` skill folder from OpenClaw's skills directory. Sign out of the account in the built-in browser's `openclaw` profile if the session is no longer needed. Knowledge base files are local in `knowledge-base` and are not removed with the skill.
FAQ
Does the skill publish notes itself?
Not by default. The agent reaches the publishing page, fills in the cover, title, text and topics and stops before the button. If confirmation is needed, it sends a screenshot to Feishu and waits for your decision.
How does the skill answer comments?
It first checks the notifications page and reports the number of new comments, a short digest of the latest 3 to 5 and signals such as abuse and suspicious links. A reply is sent after the user decides, one at a time, with the addressee name checked in the field placeholder and the Send button used instead of Enter.
What are the note replication modes?
For a note by link there are `style-only` (default), `medium` and `tight`. The first keeps the topic and engagement mechanic while the cover takes only style, color and hierarchy, with no specific elements. `tight` is enabled on direct request. The skill forbids verbatim copying and reuse of the original image.
What is the knowledge base?
A `knowledge-base` folder with an entry file `README.md` and subfolders `accounts`, `topics`, `patterns`, `actions` and `reviews`. Before a task the agent looks for entries on the topic there, and afterwards saves the conclusion, evidence, risks and next step. Entries stay local by default and are listed in `.gitignore`.
How is the skill installed?
Two ways: ask OpenClaw or Codex to install the repository by link, or run `clawhub install xiaohongshu-ops`. After installing you need to sign in to Xiaohongshu in the built-in browser.
Discussion
No questions or reviews yet.
Related
SkillsA set of 131 redfoxdata skills for Chinese social data and content: search, account analysis, video downloads, titles and banned-word checks
Self-hosted MCP server for xiaohongshu: publish notes and videos, search, read feeds, comment, like and favorite from an agent
An X (Twitter) research skill: the agent searches posts, reads threads and profiles, and assembles a briefing with source links
A social content skill: learns your voice, plans content, writes posts and, after confirmation, publishes through your logged-in Chrome
A visual-director skill for Xiaohongshu carousels: style choice, page structure, image prompts and the post caption
A collection of open Agent Skills by Huashu for content work: slides, design, data processing, video scripts and text proofreading
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| xiaohongshu-ops (this entry) | no | 2,443 | ||
| xiaohongshu-mcp | no | not required | 16,449 | a local MCP server with Xiaohongshu tools: publishing, search, feed, likes and comments from any agent with MCP |
| XHS Visual Director | no | 1,424 | a visual-director skill for image posts: style, sheet structure and prompts, with no account control | |
| RedNote-MCP | no | not required | 1,124 | an MCP server only for searching notes and reading them by link |
README in short
The description in Chinese lists the capabilities: feed and account analysis, topic ideation, a knowledge base, auto-publishing, auto-replies, downloading a note by link and replicating a note. Each capability has a sample request and a short test result, followed by installation options and the repository layout. The repository holds references on note evaluation, publishing, comments and working rules, a sample scenario for a TV drama account, reply examples and demo images.
README badge
Original SKILL.md
--- name: xiaohongshu-ops description: "小红书相关操作，覆盖账号定位、选题研究、内容生产、发布执行与复盘修复的小红书全链路运营技能。凡是小红书的浏览/搜索/发布/评论任务，默认必须使用 OpenClaw 内置浏览器流程并指定 profile=\"openclaw\"；除非用户明确要求，否则不要使用系统 open 或外部浏览器。" --- # Openclaw 小红书运营技能（通用版） 目标：构建可复用的“小红书运营”流程，让任何账号类型都能复用同一套动作框架。 ## 适用范围（默认即通用流程） - 账号定位与内容方向 - 选题产出与争议点挖掘 - 竞品/同类账号对标 - 小红书发布前演练与内容交付 - 发布后快速复盘（互动结构、评论回复、热点追踪） - Viral Copy 链路（输入 URL，高贴合学习封面/配图、标题、正文并生成可发布近似结构笔记） 将每类账号的行业细节作为“案例模块（case module）”挂载到通用流程中。 ## 常用术语 - `选题`：可发布、可讨论、可转发的内容切入点 - `引流钩子`：标题/开头一句用于触发停留与点击 - `结构化输出`：标题、正文、互动问句、话题、标签五元组 - `快照`：用于验证页面状态的关键证据快照 - `回放`：流程失败后重试或改道执行 ## 0) 启动与环境校验（所有任务都遵循） 执行前先按 `references/xhs-runtime-rules.md` 中“运行规则”执行，优先遵循失败可复用顺序。 - 固定使用内置浏览器 profile：`openclaw`，出现通道异常先切回后再重试。 - 若 browser（openclaw-manager）能力处于 disabled/不可用：先执行一次轻量重试（如 status/profiles），仍不可用则进入故障引导，明确告知用户“当前浏览器工具未启用”，并引导用户按文档启用后再继续（参考：`https://docs.openclaw.ai/tools/browser`）。 - 以 `evaluate` 为先，关键节点少量 `snapshot`，单步动作最多重试一次。 - 失败后保留已获结果，切稳健路径并汇报。 ## 1) 技能默认行为（所有任务都遵循） - **先读本技能目录下的 `persona.md`**（小红书平台专用人设/语气/发布与回复风格）。所有对外文案（发帖/评论回复/私信话术）都必须遵循。 - 开始新任务前，先读 `knowledge-base/README.md` 这个总览入口，再按 `references/xhs-knowledge-base.md` 的规则检索最近的同类记录；能复用的 pattern 不重复摸索。 - 优先输出可执行的 SOP 而非一次性内容稿 - 语言优先“能对话”而不是“写报告”：短句、口语、站位明确、可引导评论 - 所有输出默认保留“可追问点”，用于评论区继续延展 ## 2) 账号定位（可复用） 每个账号先确认 4 个变量： - 目标用户：年龄/场景/痛点（如「下班后碎片时间」「追星讨论人群」）