그림 한 장, 여러 모양
정사각, 세로, 가로 버전을 만들고 합계 가격을 먼저 알려 줘요.
쓰는 법
어시스턴트의 스킬 폴더에 스킬을 복사하세요. Claude Code라면 ~/.claude/skills/<이름>/SKILL.md예요:
mkdir -p ~/.claude/skills/pickshot-ad-sizes
curl -L https://pickshot.ai/developers/skills/pickshot-ad-sizes/SKILL.md \
-o ~/.claude/skills/pickshot-ad-sizes/SKILL.mdSKILL.md 전체
---
name: pickshot-ad-sizes
description: Re-make one picture in several shapes (square, tall, wide) on the user's Pickshot account, so one image can be used for a feed post, a story and a banner, with the price of all of them shown first. Use it when the user has one picture and needs it in more than one aspect ratio.
---
# One picture, several shapes
This skill needs the Pickshot MCP server (`get_balance`, `list_models`, `estimate_cost`, `create_image`, `get_job`). If they are missing, tell the user to connect it first (https://pickshot.ai/en/developers/mcp).
The spending rules of the `pickshot-generate` skill apply: quote first, wait for yes, cap the price, one key per request. Here there are several requests, so quote all of them and ask once for the total.
## What this does and does not do
The picture is re-drawn by a picture-editing model in the new shape. It is a new picture in the same look, not a crop: faces, text and small details can change. Any text in the picture should be checked, and logos are not guaranteed to survive. Say this to the user before you start.
## Steps
1. Get the picture as a public https link and ask where it will be used. Suggest shapes from the use: `1:1` for a feed grid, `4:5` or `3:4` for a tall feed post, `9:16` for a story or short clip cover, `16:9` for a banner or a video thumbnail.
2. Call `list_models` with `kind` `image`. Keep models that are `available` with the `edit` mode. Read the limits for `edit`: the `ratios` the model offers and how many reference pictures it takes. Choose the cheapest model that offers every shape the user needs; if no model offers one of them, say which shape is missing.
3. For each shape, write the request: `model`, `mode` `edit`, `images` set to the picture link, `ratio`, and a `prompt` that says to keep the same subject, colours and style and to recompose it for the new shape, with any text kept exactly as it is.
4. Call `estimate_cost` with `kind` `images` for each request. Add the credits and tell the user the total, one line per shape. Wait for yes.
5. Create them one by one with `create_image`, each with `max_credits` equal to its own quote and its own `idempotency_key`. Do not start the next until the last has answered.
6. Give the user every link with its shape. Point out anything that changed (text, faces) and offer a retry for that one shape, with its price.