sdt-skills
15 skills for Codex that cover Xiaohongshu content from account research and topic choice to title, draft and pre-publish QA
- 281 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The research skill runs an external command that reads public Xiaohongshu pages through a wrapper with an allow-list and a timeout
- Skills write account profiles, research exports and publication archives to disk in your working folder
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/sdt-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The research skill runs an external command that reads public Xiaohongshu pages through a wrapper with an allow-list and a timeout
- Skills write account profiles, research exports and publication archives to disk in your working folder
This is third-party code. Review the repository files before installing.
What it does
sdt-skills splits Xiaohongshu content work into fifteen separate skills called one at a time. The entry skill `sdt` decides what you need now and opens one module. If you ask only for a title, the title skill runs without research. The full ten-stage pipeline starts only on an explicit request. The skills come in layers. For context there is `sdt-account` with account facts, audience and tone limits. The evidence layer consists of `sdt-research` (read-only collection of public data with sources), `sdt-benchmark` (selecting samples free of celebrity effects) and `sdt-spread` (decoding how content spreads). Strategy is `sdt-topic`: 20 candidates, scoring, a top five. Then come `sdt-content`, `sdt-xhs-title`, `sdt-cover`, `sdt-hook`, three checks (`sdt-script-flow`, `sdt-resonate`, `sdt-ai-check`), the `sdt-wechat-html` converter for WeChat publishing and `sdt-content-system` for the archive. Repeatable operations live in five Python scripts: normalizing metrics like `3.2万`, scoring topics, checking titles, rendering HTML and a safe wrapper for reading public pages. The repository check `python tools/validate_repo.py` runs on every push.
Who it is for. Creators and marketers who run a Xiaohongshu or WeChat account and want to break post production into checkable steps.
Good fit when
- You need to pick a topic and find samples that can be repeated without a celebrity's capital or rare resources
- The title is weak and the rest of the text should stay: `sdt-xhs-title` alone is enough
- Before publishing you want the text checked for AI traces and mismatch with the account's tone
- Approved Markdown has to become inline-styled HTML for WeChat
- After publishing, the data and the original hypothesis must go into an archive so the review is honest
Not a fit when
- You need skills for other platforms such as Instagram or YouTube: the sets are tuned for Xiaohongshu
- You want autopublishing: the research tools do not publish, like or comment
- You want to copy someone else's texts and stories: the skill rules forbid carrying over wording and personal history
- You have no account and no facts about it: first-person writing needs a profile
Example request
Use sdt-xhs-title to suggest five titles for a home renovation note and check the length of eachLimitations
- The skills and README are written in Chinese and English. The agent answers in the language of your request.
- Installation is described only for Codex through skill-installer or by copying folders into `~/.codex/skills/`. For other agents you choose the skills directory yourself.
- The research skill only reads public Xiaohongshu data and does not work with a login: sign-in, likes, comments, publishing and credential export are blocked by the guard.
- To qualify a benchmark video the skill sets a threshold of at least 3,000 visible likes.
How to disable. Delete the folders whose names start with `sdt` from `~/.codex/skills/` and restart Codex. Files the skills created in your working folder, such as the content archive and account profiles, sit separately and are deleted by hand.
FAQ
How do I install sdt-skills in Codex?
Send Codex the repository link and ask it to use skill-installer to install every skill under `skills/` whose folder name starts with `sdt`. Manually, copy the folders into `~/.codex/skills/`. The skills work from the next message.
Do I need to run all 15 skills in a row?
No. By default one module runs per request: only a title, only an opening check, only topic search. The full pipeline starts when you explicitly ask for everything from start to finish.
What does the skill do with facts that are not in the data?
Missing items are marked for confirmation. The skill does not invent account experience, numbers or cases, and confirmed facts, evidence-based inferences and open questions are stored separately.
Can the skill publish posts to Xiaohongshu?
No. The research tool reads public data only: likes, saves, comments, publishing, deletion and login export are closed to it.
How does this relate to WeChat publishing?
The `sdt-wechat-html` skill turns checked Markdown into inline-styled HTML suitable for the WeChat editor. Rendering is reproducible across style presets.
Discussion
No questions or reviews yet.
Related
SkillsAn OpenClaw skill that runs a Xiaohongshu account through the built-in browser: feed analysis, ideas, note drafting, comment replies and a knowledge base
Ten skills run social media content from brief and strategy through platform copy and short video to publishing and review
The xhs command-line tool for Xiaohongshu with a ready skill: the agent searches notes, reads comments and posts as you.
A skill that checks a text before posting to Douyin, Xiaohongshu and WeChat Channels: what breaks the rules, which rule it rests on, and how to fix it
A WeChat article layout skill: turns Markdown into paste-ready HTML with inlined styles that survives pasting into the official account editor
A visual-director skill for Xiaohongshu carousels: style choice, page structure, image prompts and the post caption
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| sdt-skills (this card) | no | 281 | |
| xiaohongshu-ops | no | 2,501 | an OpenClaw skill that runs the account through a built-in browser, while sdt-skills only read public data and prepare material |
| xiaohongshu-cli | no | 2,722 | a CLI for working with Xiaohongshu, while sdt-skills add topics, drafting and checks on top of research |
README in short
The README is bilingual. It first explains why a long prompt is split into modules: a weak title should not rerun the research, and an error can be sent back to the right stage. Then come a module diagram, a table of fifteen skills with their outputs, ten workflow stages with return paths, five Python tools, tests and privacy boundaries.
README badge
Original SKILL.md
--- name: sdt description: "SDT 内容生产工具箱的总入口：根据当前任务选择最合适的 sdt-* 模块，既能完成标题、开头等单一步骤，也能组织从账号研究到发布的完整流程。用户不知道该用哪个 SDT Skill，或希望一站式完成内容生产时使用。 English: Main entry point for the SDT content-production toolkit. Use when the user is unsure which sdt-* skill to choose, wants a single step such as a title or hook, or wants an end-to-end workflow from account research to publishing." --- # SDT：内容生产系统 把账号事实、公开内容证据和 SDT 自有方法组合成可追溯的内容工具箱。每个 sdt-* 都是独立 Skill；总入口只负责识别当前任务和组织必要衔接。 ## 默认行为：单步骤优先 1. 读取当前对话，识别用户此刻要的一个交付。 2. 能直接确定模块时，只执行该模块，不要求补做上游全流程。 3. 只索取完成本步骤不可缺少的信息；不要为了“流程完整”索取无关材料。 4. 用户说“只要标题”“只检查开头”“只帮我找选题”时，在对应结果交付后停止。 5. 只有用户明确说“完整生产、从头到尾、整套做完”时，才进入完整生产线。 读 references/module-map.md 选择模块。 ## 核心规则 1. 需要第一人称代写或账号适配时先确认账号事实；纯标题公式、通用质检等单项任务不强制建完整档案。 2. 先保留来源证据，再讨论为什么火。正式对标视频的可见点赞数必须大于等于 3,000，但不可用这一绝对值代替相对表现判定。 3. 迁移传播机制，不搬运原句、事实、人物经历或独特资源。 4. 内容即使删除产品信息也必须值得看；产品只作为自然落点。 5. 任何数字、经历、客户案例和结果承诺都必须能追溯到来源。 6. 标题和开头必须在正文方向确定后制作。 7. 所有模块使用 SDT 自有规则，不依赖其他技能才能运行。 ## 运行模式 - 单项任务：默认模式，只运行标题、开头、封面、检索、选题、诊断或归档中的一个。 - 完整生产：仅在用户明确要求时，从目标账号和研究范围开始，交付一套可发布内容。 - 选题研究：停在选题池与评分，不写正文。 - 单项生产：标题、封面、开头、逐字稿等已有上游材料时单独处理。 - 复盘沉淀：把发布数据与结论写回内容资产系统。 ## 完整生产线（仅显式触发） 按以下闸门顺序执行。读 references/pipeline.md 获取每一阶段输入、输出和退回条件。 1. 使用 sdt-account 建立账号档案。 2. 使用 sdt-research 获取账号、笔记、标题、封面和可见指标。 3. 使用 sdt-benchmark 筛出垂直、邻近、跨域三类样本。 4. 使用 sdt-spread 提炼爆款的承重传播机制。