Multiplatform Content Production Workflow
Run a complete Chinese content workflow from one brief, topic, or outline to audited deliverables for five publishing platforms.
Basic info
- Name
- Multiplatform Content Production Workflow
- Description
- 从一句话创作简述、明确题目或一份 Markdown 大纲自动编排中文多平台内容生产:完成选题与调研、证据化大纲、A/B 母稿、公众号/小红书/知乎/微博/头条十份终稿、34 个标题、五套平台配图、独立公众号封面、图片优化、公众号 HTML 和最终 QA。用于完整写作流水线、多平台内容矩阵、A/B 文风实验或一键内容生产;默认自主运行并保留可审计决策,不发布、不登录平台。Do NOT trigger when 用户只需要单篇轻量润色、只起标题、只配图、只排版、内容发布或发布后运营。
Multiplatform Content Production Workflow
Run an audited local workflow for five-platform content.
When to use it
Start from a brief
I have only a one-line creative direction and want the workflow to choose a viable topic, research it, and carry it through every production stage.
Use a chosen topic
I already know the subject and need two master variants plus finished versions for WeChat, Xiaohongshu, Zhihu, Weibo, and Toutiao.
Use an approved outline
I have a Markdown outline and want it verified, normalized, and expanded into a complete package with titles, illustrations, a WeChat cover, and clean HTML.
Won't take
This is not for a quick edit, one title, one illustration, layout-only work, publishing, scheduling, or post-publication analytics.
What it produces
The endpoint is a verified local delivery package, not a published post or platform draft.
- Writing set: Creates two platform-neutral masters and ten reviewed platform versions
- Title set: Produces exactly 34 candidates and one selected title for each platform
- Visual set: Produces five illustration suites plus one exact-size WeChat cover
- Publishing package: Builds platform folders, optimized images, WeChat clean HTML, and a browser preview
- Audit trail: Preserves versioned decisions, hashes, gates, QA records, and recovery state under
runs/<run-id>/ - External activity: May research public sources and call validated local providers; all deliverables stay local
- Never: Logs into platforms, creates platform drafts, schedules posts, publishes, or reads performance data
Visual V3 runs body illustrations and the WeChat cover as independent lifecycle components. Body work uses a five-platform Canary barrier, bounded concurrency, per-image retries, exact approved text, and hash-based PASS reuse; cover failures do not discard completed body pixels. Persisted artifacts use run-relative paths or PortablePathRef values and never retain resolved local user paths or credentials.
Prerequisites & boundaries
Prerequisites
Provide exactly one starting point: a brief, a chosen topic, a Markdown outline path, or pasted outline text. All required local provider skills and their matching contracts must be available.
Neighbor skill split
| Need | Hand off to |
|---|---|
| Research only | collect-sources |
| Drafting only | draft-content |
| Illustrations only | post-illustration-images |
Scenarios it declines
- Lightweight editing or a single-platform writing request
- A request for only titles, images, layout, or proofreading
- Any publishing, scheduling, account login, or post-publication operation
Subtle edges
- Autonomous mode records and approves fixed-rule gates without pretending the user approved them
- Reviewed mode pauses only at the named decision gates
- A failed provider contract blocks initialization instead of being replaced by a generic agent
Version info
Local public Skill catalog snapshot, showing only public-safe fields.
Skill files
(258)SKILL.md
SKILL.md · Markdown