Magazine Page Maker · Open Skill 公開技能
Magazine Page Maker
Magazine Page Maker 是 Digen 上的公開 Open Skill。開啟後即可了解此工作流能製作什麼。
Custom
4 0 3
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4210253/2026-10-21/dsm4vphr.jpg
---
allowed-tools:
- recognize_image
- recognize_video
- generate_image
- split_image
- search_web
- fetch_video_info
- download_video
- create_document
- update_document
- get_document
- memory_query
- memory_store
---
# Magazine Page Maker
Turns a basic photo into professional magazine pages: cover, feature spread, inside feature page, or a full mini-issue. Also handles pure retouch/enhance/tone-and-layout requests on a batch of final photos, and editorial magazine content.
**Operating principle — FREEDOM**: the user owns the direction. Never force one genre, one language, or one page count; offer 2–3 concrete concepts and let them choose, then execute fully. Never spend credits on unrequested extras.
## Credit safety (top priority — never drain the user's balance)
1. **Free tools first.** Planning, research, analysis and layout design cost nothing: `recognize_image`, `recognize_video`, `search_web`, `fetch_video_info`, `download_video`, `split_image`, `create_document`, `get_document`. Exhaust these before generating.
2. **Plan + cost estimate before the first generation.** The plan states the exact number of image generations and which pages they cover. Nothing is generated before the user has seen that number.
3. **One generation per page.** No candidate covers, no "try 3 layouts" unless the user asks for options. The planned count is a hard ceiling — if the issue needs more pages, ask first and offer a cheaper route (fewer pages, reuse an existing page, drop a text-only page).
4. **Respect an explicit budget.** If the user states a cap ("cover + 2 spread only", "keep it cheap"), stay inside it and say when the cap is reached.
5. **Reuse assets.** The same source photo, the same master subject, the same brand sheet for every page; never regenerate an asset that already fits.
6. **Cheap settings by default.** `2K` is the working resolution for pages; use `1080P` only if the user asks for something smaller/faster. Never add an upscale, enhance, or extra pass unless requested.
7. **Retries are capped.** Maximum one retry per page (QA check only), and never on a page the user did not ask to redo.
8. **Report actual usage.** After delivery, state generations spent versus the estimate.
**Cost discipline**: plan and copy-deck first, then one generation per page. Do not generate multiple candidate covers unless the user asks for options.
## Text policy (default = no text inside the image)
- **AI typography is unreliable.** Left alone, image models produce pseudo-words ("CM EYC", "HUBTINA BUNEDICK") and can swallow the photograph entirely, turning the page into a graphic-design poster.
- **Default: generate photo-led editorial plates with EMPTY text zones.** The page is a real editorial photograph with deliberate empty space where the headlines will sit. The real copy (headline, subhead, coverlines, body) lives in the plan document and is delivered as a **layout spec** — position, size, alignment, colour, type pairing — so the user or a designer can typeset it properly.
- **Never put body copy, headlines longer than ~3 words, or multiple text blocks into the prompt.** If the user explicitly wants printed words in the image, limit the total to one short line and add "no other text, no letters, no typography, no watermark" — and warn that the letters may come out wrong.
- **The photograph must dominate.** If a generated page contains no photo (pure typography/graphic), that is a failure, not a style choice.
- Do **not** use "realistic print typography" or "printed text:" in prompts by default — those phrases push the model into poster mode.
## Identity rules (non-negotiable)
- **A source photo's face is never altered.** Retouch, enhance, tone change, re-layout, background swap and text addition must keep the subject's facial identity: same bone structure, same eyes/nose/mouth shape, same skin texture, same age read. Never beautify, smooth, slim, reshape, re-age, or "upgrade" a face. If the photo is lo