演示文稿
Compose Image
User wants to compose, merge, collage, or restyle selected/attached images into one new picture, or generate a picture from a prompt. Use when the outcome is a single durable image under /artifacts. Do NOT trigger for editable slides or a 海报 (slides / poster).
Player-YN · 已有 AI 审核记录
安装与使用
运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。
npx skills add Player-YN/PawWork_ZhuaZhua --skill Compose Image技能说明
name: Compose Image description: User wants to compose, merge, collage, or restyle selected/attached images into one new picture, or generate a picture from a prompt. Use when the outcome is a single durable image under /artifacts. Do NOT trigger for editable slides or a 海报 (slides / poster).
Compose Image playbook
This skill produces one raster PNG under /artifacts. It is not how you build editable slides or a 海报. Those are tldraw canvases — skill slides / poster (native icon / motif / chart / image slots). Do not flatten a deck into an image to "finish" it. Do not use this skill to fill a slide or poster visual slot; those skills derive deck act=read catalog="image-brief" then acquire, then attach the durable path.
- Bind the user's selected images to this session first (host/UI). Do not invent group writes.
inspectbound image items only if you need to confirm which references exist.- Call acquire with
action: "image"and a concreteprompt. - Pass
itemIdsof bound image WebItems for compose / image-to-image. OmititemIdsto use every bound session image. - This is the only place where lettering inside the picture is legitimate: when the user explicitly wants text baked into the fused image, pass
allowText: trueto acquire. Everywhere else the host stamps image prompts no-text — captions belong in canvas text nodes. - If Settings has not enabled image generation, tell the user to open ⋯ → 设置 and tap OpenRouter 生图. Do not fake pixels in
run. - The host writes a real PNG under
/artifactsand validates bytes. Observation is short JSON (path, artifactId) — never paste base64 into chat.