| name | social-cover-director |
| description | Use when the user wants to plan or generate a social media cover image for a Xiaohongshu note, Douyin/TikTok photo post, Reels/Shorts cover, WeChat/Video Account post, Bilibili/Weibo post, Pinterest pin, Instagram cover, Facebook/Meta social post, or similar creator/social content. Handles platform-native cover routing, aspect ratio, scroll-stopping hook, exact on-image copy, reference-image roles, and image.generate prompt construction for one cover or a small variant set. |
| allowedRuntimeModes | ["chatroom","redclaw","image-generation"] |
| allowedTools | ["workflow"] |
| activationScope | turn |
| autoActivate | false |
| activationHint | 当用户要做小红书封面、笔记封面、社媒封面、视频封面、图文首图、Reels/TikTok/Shorts 封面、朋友圈/视频号/微博/B站/Instagram/Pinterest 等内容封面时,调用 `Operate(resource="skills", operation="invoke", input={ "name": "social-cover-director" })`。这是单张封面或少量封面变体技能;成套卡片、轮播图、商品详情图继续优先用 image-director。 |
| contextNote | social-cover-director 只负责社交媒体内容封面的点击钩子、封面文案、画面策略和生成提示词。不要把它扩成整套图文卡片规划;不要把封面制作降级成普通配图或商品白底图。 |
| hookMode | inline |
Social Cover Director
Use this skill to make one social-media cover image, or a small set of cover variants, for creator content.
It adapts the social-promo planning flow to RedBox:
content/topic + platform/surface
-> reader promise + scroll-stopping hook
-> cover type + aspect ratio
-> attention mechanic + visual style pack
-> exact on-image copy
-> reference-image role map
-> image.generate prompt
-> text/identity/social-native QA
Core Stance
- The cover must stop the scroll before it explains the whole content.
- Prefer a strong human-readable promise, contrast, clear subject, clean title zone, and platform-native composition.
- Do not make a catalog poster unless the user explicitly wants a product-commerce cover.
- Do not force a product, person, or logo into the cover if the stronger hook is a mood, problem scene, cultural moment, checklist, or visual metaphor.
- Exact on-image copy is a hard boundary. Use only confirmed cover text; do not let the image model invent extra labels.
- If the user wants a full carousel, multi-card note, ecommerce image set, or detail-page images, route to
image-director instead.
Required Inputs
Useful inputs:
- Content source: draft note, topic, title, outline, pasted article, video idea, product facts, or campaign brief.
- Platform/surface: 小红书笔记, 抖音图文, Douyin/TikTok/Reels/Shorts cover, 微信朋友圈, 视频号, 微博, B站动态, Instagram, Pinterest, Facebook/Meta.
- Market and on-image language, when not obvious.
- Reference images, when available.
If platform is missing and the content is Chinese creator content, default to 小红书笔记. If content itself is missing, ask for the topic/draft first. If only market/language is missing, infer from the user language unless the final usage seems overseas.
Reference Loading
These references are bundled with this skill and may be used directly:
references/cover-routing.md: platform, surface, aspect ratio, and variant count.
references/attention-design-router.md: attention mechanics and style packs.
references/image-generate-prompt-contract.md: RedBox image.generate prompt contract.
Workflow
- Extract the content promise:
- target reader
- core tension or curiosity
- desired click reason
- one thing the cover must make obvious
- Route platform and surface.
- Pick cover type: