Turn the finished draft produced by compose into note copy suitable for rednote (Xiaohongshu). By default, deliver Markdown text in the reply for the client to copy and publish manually; this skill does not call the rednote API, does not send automatically, and does not modify the original draft.
Client-readable description
What this skill can do
Distill a clear sharing angle from the article and organize it into a note suited for quick mobile reading: a title with an informational promise, 3–5 short paragraphs, relevant topic tags, and image suggestions that match the content.
| What the client wants | What the skill does | What the client sees |
|---|---|---|
| Post the article as a Xiaohongshu note | Distill the angle, rewrite the title and short paragraphs, add tags | A copy-ready rednote Markdown text |
| Need direction for visual assets | Provide suggestions for the subject, composition, text, and aspect ratio of the cover/accompanying images | Actionable image suggestions; can connect to soia-pkm-cover-image |
| Publish to rednote | Only generate publishing text | "Produce text, publish manually"; will not call an API or publish content |
How the client uses it
- Say "post to Xiaohongshu", "Xiaohongshu note", or "rednote this one", and provide the finished draft, file content, or path.
- If requested, also specify the target audience, account tone, prohibited title words, whether to highlight methods/checklists/stories, and the desired image style.
- The Agent first determines a single sharing angle, then outputs the title, body, tags, and image suggestions; by default it does not overwrite the original draft, and only saves separately when the client specifies a path.
- The client manually copies the copy, prepares images, and publishes to rednote; this skill does not perform platform backend operations on their behalf.
Dependencies and installation
Install this skill (single skill):
npx skills add soia-team/soia-open-skills -g -a '*' -s soia-pkm-publish-rednote-card -y
- This skill is a pure LLM rewriting process, with no scripts, no private configuration, no account credentials, and no rednote API dependency.
soia-pkm-cover-imageis an optional connecting skill: if you need to generate a cover bitmap, install and run it first; this skill only provides image direction and does not generate or upload images for the client.soia-pkm-compose-article-draftis a common upstream artifact, but it is not a hard installation-level dependency; you can also directly provide any finished draft.- Currently not connected to the rednote API. Any "publish" only means generating text; manual publishing is done by the client.
Logs and completion receipt
Each execution must report the actual processing scope, the number of titles/paragraphs/tags, and the manual publishing boundary, and must not describe "generated" as "published". Minimum format:
完成:已将 <输入范围> 改写为 rednote 笔记文本,未调用平台 API。
日志摘要:
- started: <输入来源、目标读者与分享角度>
- processed: <标题 1 个、正文 3–5 段、标签数量、配图建议数量>
- created/updated: <回复文本或客户指定的输出位置>
- skipped/failed: <未处理内容及原因;没有则写“无”>
验证:
- <标题吸引力与事实一致性、段落数量、标签相关性、配图建议核对结果>
问题与下一步:
- 请客户人工复制文案并准备图片后发布;<其它问题,没有则写“无”>
Rewriting rules
1. First define one sharing angle
- Choose an angle from the original draft that readers can immediately understand: solving a problem, avoiding a pitfall, reviewing an experience, or providing a set of methods.
- The title, body, tags, and images must revolve around the same angle; do not compress the entire article into a list of unrelated viewpoints.
- Preserve the original draft's facts, qualifying conditions, and verifiable sources; do not add numbers, cases, citations, or effect promises unsupported by the input.
- If the original draft has multiple equally important themes, first ask the client to choose; when risk is low, choose the clearest main line and explain it in the reply.
2. Title
- Generate 1 main title, and if necessary attach 2 alternative titles for the client to choose from; the main title should have a specific information promise or scenario conflict, and avoid vague "value-sharing".
- A small number of semantically relevant emoji may be used, but emoji cannot replace keywords, nor can exclamation marks be stacked, benefits exaggerated, or false authority used.
- The title accurately reflects the body and does not make clickbait-style promises of results the body does not provide; comply with the platform prohibited words and brand restrictions provided by the client.
3. Body and tags
- Keep the body to 3–5 short paragraphs, with each paragraph expressing only one information point; state the scenario/pain point as early as possible at the beginning, provide methods or evidence in the middle, and conclude and guide interaction at the end.
- Use short sentences that are easy to read on mobile, moderate line breaks, and specific examples; retain necessary technical terms, but explain terms the first time they appear.
- Place topic tags after the body, usually giving 5–10 tags directly related to the topic; narrow progressively from broad topics to specific issues, and do not add irrelevant trending tags to chase traffic.
- If the user has not provided an account tone, adopt a sincere, specific, and verifiable sharing tone; do not impersonate personal experience, and do not claim to have personally used products or methods that were not provided.
4. Image suggestions
- Provide at least 1 cover suggestion, explaining the main subject, composition, primary color, text hierarchy, and recommended ratio; suggestions may be made according to the 1:1 Xiaohongshu square image scenario.
- When the article has steps, comparisons, or a checklist, add 1–3 inner-page card suggestions, specifying the single piece of information each card carries.
- The titles, numbers, and sources in the suggestions must be consistent with the copy; do not put unverified facts into image text.
- When the client needs an actual cover, explain that the suggestions can be given to
soia-pkm-cover-imagebut do not claim that the image has been generated, uploaded, or published.
Output format
The default output is Markdown:
## 标题
<吸睛标题,可带少量 emoji>
## 正文
<第 1 段>
<第 2 段>
<第 3–5 段>
## 话题标签
#主题标签 #具体问题 #方法标签
## 配图建议
- 封面:<主体 / 构图 / 配色 / 文字 / 比例>
- 内页 1:<单一信息点;没有则省略>
Outside the list, notes on factual trade-offs, prohibited words, or image connection may be attached. When the client has not specified a save location, only produce reply text; when a location is specified, then write it into a separate draft file, and never overwrite the source file. Do not output fake "published" links or API results.