安装与下载
1 ClawHub CLI
从源直接安装技能的最快方式。
npx clawhub@latest install jishudb-xiaohongshu-content-factory
2 手动安装
将技能文件夹复制到以下位置之一
~/.openclaw/skills/
{'/skills/'}
优先级:工作区 › 全局 › 内置
3 提示词安装
将此提示词复制到 Openclaw 即可自动安装。
请帮我使用 Clawhub 安装 skill-jishudb-xiaohongshu-content-factory。如果尚未安装 Clawhub,请先安装(npm i -g clawhub)。
更倾向于下载?
获取包含原生技能文件的ZIP压缩包。
基本信息
jishudb-xiaohongshu-content-factory
安装指南
- ClawHub CLI
从源直接安装技能的最快方式。
npx clawhub@latest install jishudb-xiaohongshu-content-factory - 手动安装
将技能文件夹复制到以下位置之一
全局模式
~/.openclaw/skills/
工作区
/skills/
优先级:工作区 > 本地 > 内置 - 提示词安装
将此提示词复制到 OpenClaw 即可自动安装。
请帮我使用 Clawhub 安装 jishudb-xiaohongshu-content-factory。如果尚未安装 Clawhub,请先安装(npm i -g clawhub)。
更倾向于下载?
获取包含原生技能文件的ZIP压缩包。
下载技能 (.zip)
什么是
Xiaohongshu Content Factory is an Openclaw Skills workflow for producing a ready-to-review content pack—not just a list of prompts or titles. It develops distinct post angles, complete copy, title alternatives, source mappings, and finished PNG or JPEG covers, typically in portrait 3:4 format at 1080 × 1440 pixels.
The packaged workflow uses JishuDB to retain research, brand preferences, drafts, final copy, citations, and revision history. It supports human review and handoff, but does not publish or schedule posts, crawl protected Xiaohongshu content, or guarantee reach.
应用场景
Create complete Xiaohongshu posts and usable cover files from a topic, industry, or product.
Develop multiple content angles and title options for a campaign or product launch.
Turn public, source-grounded research into audience-focused posts with claim-to-source mappings.
Produce content packs that need PNG/JPEG cover exports rather than image-generation prompts alone.
Maintain brand preferences, prior content, and revisions in an authorized JishuDB knowledge base.
Prepare deliverables for a person to review and publish independently.
工作原理
Confirm the requested post count, language, output location, JishuDB access and write scope, approved destination, and availability of image or graphic export. The default is one post and one cover.
Assign a filesystem-safe run ID and retrieve relevant brand preferences and prior content from the selected knowledge base. Review the full content inventory when checking for previously used angles.
Research relevant public facts and audience questions using authorized sources. Record source provenance and coverage; avoid protected-platform crawling and unsupported claims.
Select a useful angle for each post, then write the intended audience, evidence-backed promise, title alternatives, selected title, hook, complete body, and natural closing action.
Refine the draft for clarity, usefulness, source accuracy, and brand consistency. Render one finished cover per post, export it as PNG or JPEG, and inspect the actual file for legibility, dimensions, contrast, clipping, and consistency with the post.
Create the content-pack files: posts.md, cover-index.md, sources.md, and revision.md. Save matching final copy and map each post to its real cover filename and asset rights.
Persist required source and content records in JishuDB through the host's MCP tools. Verify jobs, document text, knowledge-base membership, current revisions, and searchability before reporting completion.
Hand off the actual output paths, selected knowledge-base ID, saved record IDs, coverage limitations, and a truthful status: COMPLETE, PARTIAL, BLOCKED, USER_ACTION_REQUIRED, or FAILED. Publishing is not part of the workflow.
配置指南
This skill requires an Agent Skills-compatible host, JishuDB MCP access with the necessary read/write permissions, authorized research access, approved local file output, and an image or graphic tool that can export usable cover files. JishuDB setup guidance is bundled with the skill. No CLI installation command is specified in the provided skill file.
Install or load the skill in a compatible host and complete the bundled JishuDB setup using references/jishudb-setup/setup.md if needed.
Check JishuDB capabilities with kb_get_capabilities({}); inspect relevant live schemas, write support, limits, and upload configuration. Read kb_read_handbook({}) only when behavior is unfamiliar.
Call kb_list({}) and use a returned knowledge-base ID for the previously authorized destination. If a new knowledge base is necessary and authorized, create it with kb_create and verify it with kb_list.
Confirm that the required output and cover-rendering capabilities work. Follow references/cover-rendering.md and its checked local-path probe before exploring other tools.
Choose the user's output directory or an approved host workspace location. The workflow creates a new xiaohongshu-content-factory- directory and preserves existing files.
No CLI installation or configuration command is specified by this skill.
Xiaohongshu Content Factory 数据架构与分类体系
The workflow organizes deliverables in a per-run directory such as xiaohongshu-content-factory-/ and retains stable run IDs, revision markers, record UUIDs, timestamps, and retry keys.
File or record Contents and purpose
posts.md Finished post copy, including selected titles and complete bodies.
covers/ Actual exported, nonempty PNG/JPEG cover files; editable sources are included when supported.
cover-index.md Mapping between posts and real cover filenames, with asset-rights information.
sources.md Claim IDs mapped to supporting passages and source details.
revision.md Run and skill version, revision history, predecessor references, edits, fingerprints or identifying titles, and delivery status.
Sources record Permitted retained material or limited excerpts, provenance, coverage, and source document IDs.
Brand brief record Approved preferences, audience assumptions, constraints, and selected defaults.
Angles and draft record Topic ideas, selected and rejected angles, title alternatives, first-pass posts, and claim IDs.
Final and citations record Finished copy, cover copy and relative filenames, asset attribution, and claim-to-source mappings.
Revision and usage record Run/version identifiers, predecessor IDs, edits, content identifiers, and status; publication status only when separately evidenced.
Source metadata includes title, publisher, stable public URL, publication and collection dates, coverage, and permitted supporting passages. Numerical claims should also record period, geography, units, definition, and whether values are observed or forecast. JishuDB records use stable UUIDs and RFC 3339 creation timestamps; linked notes must be read back and verified. Large records may be split into numbered parts with a part manifest to fit live limits.
高级特性
Reuses authorized JishuDB consent, selected destinations, brand preferences, and content history instead of requiring an intake form for every run.
Uses full paginated document inventories when checking prior content; search results alone are not treated as proof that an angle is new.
Tracks source provenance, claim IDs, numerical-claim context, asset attribution, revisions, and content status.
Supports JishuDB URL imports, prepared direct uploads, and supported small text uploads according to advertised capabilities and limits.
Uses idempotency keys, stable identifiers, job polling, readback checks, and reconciliation procedures to reduce duplicate writes and safely recover uncertain operations.
Preserves drafts as separate revisions and uses timestamp-guarded updates to resolve concurrent changes without blindly overwriting content.
Inspects exported cover files and requires copy-to-cover agreement; if an image model cannot render text reliably, available graphic tooling can compose the final title.
Reports partial persistence or incomplete cover work explicitly, with outstanding files or job/record IDs and a next recovery action.
Does not automatically publish or schedule posts, access protected social content, or claim guaranteed engagement or platform approval.
name: xiaohongshu-content-factory
description: >-
根据主题、行业或产品制作小红书图文内容包,适用于选题、标题、完整笔记文案
和成品封面图;交付可用 PNG/JPEG 封面,不只是提示词,不自动发布。 /
Create Xiaohongshu content packs with distinct angles, title options, complete
posts, and actual PNG/JPEG covers from a topic, industry, or product.
Use when finished posts and cover files are required; retain sources and
copy versions in JishuDB without publishing automatically.
compatibility: >-
WorkBuddy or another Agent Skills-compatible host with JishuDB MCP, permitted
web access, approved local file output, and image or graphic generation that
exports usable cover files. JishuDB setup is included in this package.