安装指南

1. ClawHub CLI 从源直接安装技能的最快方式。
   npx clawhub@latest install aikey-cover
2. 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
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 aikey-cover。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

AI KEY·封面图制作 turns a person photo, a background image, and a video topic into a coordinated set of platform-ready covers. Each language-and-ratio combination is composed independently—for example, 3:4, 4:3, 16:9, or 9:16—so layouts are adapted rather than stretched or cropped from a single finished image. Bilingual covers use separate images with consistent meaning and natural wording.
As an Openclaw Skills workflow, it also applies the cover-text rules delegated to /aikey-title, checks each image’s actual pixel dimensions, and archives the results with a production note. It creates images only: it does not edit videos, publish, or upload covers.

应用场景

Create a coordinated set of video covers for several platforms and aspect ratios from supplied person and background images.
Prepare separate Chinese and English cover versions while keeping their meaning consistent.
Produce a replacement for one missing or rejected platform size without regenerating the whole set.
Turn a video topic, script, or supplied cover text into a checked cover layout.
Archive finished cover assets and production details for later revisions.
Generate image prompts and layout plans when no suitable image-generation or editing tool is available.

工作原理

Read the required family agreement, interaction rules, skill memory, rule cards, and platform-size reference. Stop if the required agreement cannot be read; do not substitute memory for it.
Collect and inspect the person image, background image, and topic or script. Confirm the role of each image, check the video’s production progress when relevant, and ask for missing information one choice question at a time.
Delegate cover wording to /aikey-title. Apply its structure and red-line checks, respect platform text limits, and create separate, faithful language versions when bilingual output is requested.
Read the configured platforms and size table, then show the user the required language-by-ratio deliverables. Honor any platform, language, size, or image-count instructions given for the current task.
Generate each language-and-ratio cover separately with a reference-image-capable tool. Feed in both original source images for every generation, preserve the subject’s identity and set-wide visual style, and re-layout the subject and text for each canvas.
Inspect every result and verify actual pixel dimensions with a read-only image tool. Check text accuracy, composition, identity, distortion, and applicable safe-area rules; remake only failed images.
Copy verified outputs into a dated topic folder, add 制作说明.md with the wording, title structure, filenames, pixel dimensions, prompts, and tool details, then report the deliverables and any incomplete items. Do not publish or upload.

配置指南

Use this skill within an OpenClaw environment where the skill files, vault, configuration, and any image-generation or editing tools are accessible. No package installation command or external API setup is specified by the skill.
Before starting a task, read the required references and configuration:

From the skill's available workspace, inspect the referenced files:

cat aikey/references/家族公约.md
cat aikey/references/交互规范.md
cat aikey/references/规则卡.md
cat aikey/references/平台尺寸表.md
cat config.yaml
The paths above reflect the skill's documented references; resolve them against the actual skill workspace and vault. Also read the relevant skill memory under {vault}/08-技能记忆/aikey-cover/ and _通用/. If the family agreement is unavailable, explain that the red lines cannot be guaranteed and stop. Use an image tool available in the current environment; if none is available, provide per-image prompts and layout plans instead of switching to a paid interface. Read credentials from environment variables when required, and never write them to files.
AI KEY·封面图制作 数据架构与分类体系
The skill uses configuration and reference files to determine platform requirements and working rules:
Data or artifact Purpose
config.yaml → platforms Configured platforms and applicable cover-text limits
references/平台尺寸表.md Starting reference for platform aspect ratios and dimensions
references/规则卡.md Evidence-based production and validation rules
{vault}/08-技能记忆/aikey-cover/ and _通用/ Skill-specific and shared memory; check for duplicates before adding notes
{vault}/{config.paths.drafts}/封面/YYYY-MM-DD-主题/ Date-based destination for a cover set; use the user’s specified date when provided
制作说明.md Topic, localized cover text, /aikey-title structure ID, filenames, actual pixels, final prompts, tool, and whether HTML text overlay was used
Cover filenames must include language, ratio, and orientation, such as 中文-3比4-竖版.png or 中文-16比9-横版.png. Add revision identifiers before the extension, for example 中文-4比3-横版-v2.png. If a same-day topic folder already exists, create a numbered folder for a new task; for revisions to the same set, reuse the folder and add -v2 to new filenames. Do not overwrite existing files unless the user explicitly requests it. Source materials remain unchanged.

高级特性

Generates each language-and-aspect-ratio combination independently, preserving coherent visual styling without stretching a master image.
Supports bilingual sets with one language per image and faithful, natural translations.
Delegates cover copy and red-line checks to
/aikey-title
, recording the selected structure ID in the production notes and delivery report.
Adapts output to configured platforms and supports targeted fixes that regenerate only missing or failed covers.
Verifies actual image pixels with read-only tools such as
sips
or Python PIL rather than trusting filenames.
Uses an HTML text-overlay fallback if generated Chinese text is wrong twice; documents when this alternative is used.
Records prompts, tool details, dimensions, and corrections to support repeatable revisions and future learning.
Enforces a strict no-upload, no-publish boundary; completed work ends with local delivery and a clear status report.

name: aikey-cover description: | AI KEY·封面图制作 ——视频封面图制作。用你的人物素材图 + 背景图 + 主题，按要发的平台一次做齐每种比例的封面（竖版 3:4 / 横版 4:3 / 宽屏 16:9 / 长竖版 9:16，可分中英文），每个比例重新排版、不拉伸，逐张核实际像素后按日期归档。封面大字归口 /aikey-title 定稿并过红线；只做图、不发布。不负责想封面写什么字（→ aikey-title）、剪片子（→ aikey-cut）。
触发方式：/aikey-cover、/做封面、/封面图、/aikey-封面图、「做视频封面」「出一套封面」「中英文封面」「B站封面」「Instagram 封面」「各个尺寸的封面」
Video cover images. From a person photo, a background and a topic, produces one cover per aspect ratio each target platform needs (3:4 / 4:3 / 16:9 / 9:16, optionally in two languages), re-laid-out per ratio rather than stretched, each file's real pixels checked, filed by date. Cover text is defined by /aikey-title and passes the red lines. Makes images only, never publishes. Not for: deciding the cover text (aikey-title), editing the video (aikey-cut).
Trigger: /aikey-cover, "make video covers", "covers in every size", "bilingual covers"
—— AI KEY · 不给公式，给判据。每条规则都标了实测代价。
slug: aikey-cover
displayName: AI KEY·封面图制作