blueprint-magic-image-production

name

blueprint-magic-image-production

description

Produce print-safe image generation briefs, prompts, QA checklists, style guidance, and delivery guidance for the book project 左手藍圖，右手魔法. Use when Codex needs to create, review, regenerate, or specify book figures from docs/publishing image IDs, including direct AI-generated concept images, scenario illustrations, four-panel dialogue comics, HTML-to-image layouts, screenshots, Mermaid diagrams, QR/external assets, Traditional Chinese labels, 17cm x 23cm book placement, gutter-safe composition, consistent book-wide visual style, and publication-quality image output.

Blueprint Magic Image Production

Overview

Use this skill to turn one planned book figure into a production-ready image brief or finished-image QA pass. Favor book readability, print safety, Traditional Chinese correctness, and the project image inventory over decorative novelty.

Style Decision System

Choose image style from the manuscript purpose and the image type, not from one universal template. Before reviewing or regenerating a figure, decide what job the image is doing in the book, then choose the generation route that fits that image type.

Image type	Use when the manuscript needs	Preferred production route	Default style
高畫面感插圖	Big idea, emotional metaphor, product vision, chapter-opening promise, memorable scene	Direct generated image, then optional text overlay	Bright editorial illustration
Style A+ 彩色圖解版	Process, comparison, rule, relationship, safety boundary, checklist, architecture relationship, teachable concept	HTML-to-image or reliable text renderer	Color-rich information diagram
Actual screenshot	Operation steps, tool UI, terminal output, cloud settings, editor behavior, verification evidence	Real screenshot with redaction	Native tool UI
Mermaid / technical diagram	Editable flow, sequence, ERD, architecture diagram, decision tree	Mermaid export, or Style A+ redraw when raw output is too plain	Technical diagram
QR / external asset	External link, publisher-approved asset, downloadable companion material	QR/external asset pipeline	Print-safe utility graphic

Do not regenerate or reject a figure only because it is not Style A+. First compare it with the source manuscript, figure purpose, print readability, and book-wide consistency.

Aspect Ratio Decision System

Do not make every figure 3:2. Choose the ratio from the figure's job on a 17cm x 23cm book page, then follow the checklist's 建議比例 for that image ID.

Treat ratio as a design starting point, not a cage. If the figure becomes hard to read, has excessive empty space, or contains too much nonessential UI chrome, adjust the canvas to the content: enlarge type, shorten text, crop unused space, and prefer a content-driven plate size over a formally perfect ratio.

Ratio / book size	Use when	Source-pixel target
15:21 一頁滿版直式 (safe area about 15cm x 21cm)	Only when the user explicitly asks for a full-page portrait figure, or the checklist already marks the item as full-page. Good for full-page chat screenshots, one-page vertical timelines, or a long top-to-bottom flow.	3000x4200px minimum; 3600x5040px for text-heavy line art
1:1 方圖 (about 15cm x 15cm on one page)	Concept images, role maps, compact checklists, product-vision scenes, metaphor images, or diagrams that should feel like a centered single-page plate.	3000x3000px minimum
4:5 / 3:4 短直式	Short vertical flows, SOPs, document-like screenshots, conversation screenshots, or process stacks that need vertical reading but should not occupy a full page.	3000x3750px or 3000x4000px
4:3 橫式	Balanced diagrams, UI mockups, technical drawings, and screenshots that need more height than 3:2.	3200x2400px or 3600x2700px
3:2 橫式	Wide comparisons, three-stage diagrams, architecture/data-flow layouts, and figures that benefit from horizontal scanning.	3600x2400px
16:9 寬螢幕	Real desktop/browser/cloud screenshots, dashboards, DevTools, terminal/log captures, and screen-native UI where cropping would hide context.	3200x1800px or 3840x2160px

Use 15:21 full-page portrait sparingly. If the user has not explicitly requested a full-page figure, prefer 1:1, 4:3, 3:2, 16:9, 4:5, or 3:4 based on the image purpose. If a ratio starts to force tiny text, split the content into multiple figures instead of shrinking everything.

For book-facing rebuilt screenshots or operation diagrams, use 書本大字精簡操作圖版｜HTML→圖片 when a real screenshot would print poorly. These figures should use real workflow content but rebuild the visual surface with large readable type, fewer words, less whitespace, and a crop that follows the main content.

Style A+ 彩色圖解版

Use the approved Style A+ 彩色圖解版 direction for information diagrams and precise explanatory figures. The reference example is:

• assets/images/final/fig-02-008-style-a-rich-icons.png
• assets/images/source/fig-02-008-style-a-rich-icons.html

Style A+ means a clean technical-book information diagram with more visual explanation than text:

• Use strong icons, objects, line work, arrows, gates, lanes, cards, badges, and color-coded paths to carry the idea.
• Reduce visible text aggressively. Keep only the title, key technical names, short status labels, and necessary one-line explanations. Put fuller explanation in the manuscript caption or surrounding text.
• Avoid figures that are mostly text cards. A three-column layout is acceptable only when each column has a clear icon/object, status, and visual relationship.
• Use a richer but controlled palette: red for danger/blocked/secret, green for safe/pass/shareable, blue for system/rule/guard, purple for repository/knowledge structure, amber for caution or optional emphasis.
• Show the relationship visually whenever possible: blocked path, allowed path, before/after, input/output, gate, flow, ownership, or state change.
• Keep the overall feel like a polished printed technical book, not a slide deck, landing page, dark poster, or dense dashboard.

Style A+ is the default for information diagrams, comparisons, workflows, control rules, and safety concepts. It is not mandatory for 高畫面感插圖 when a more immersive image better matches the manuscript.

High-Visual Illustration Direction

For 高畫面感插圖, concept images, product vision images, and chapter-opening scenes, the approved direction is a bright editorial illustration with strong objects, concrete scenery, atmosphere, and reader memory value. Approved examples:

• assets/images/final/fig-00-002-blueprint-magic-concept.png: blueprint in one hand, AI building team / magic construction metaphor, warm desk context.
• assets/images/final/fig-00-003-habit-life-rpg-vision.png: product vision mockup with gamified UI, character, habit cards, boss, rewards, and progress path.

Use this direction when the image should make the reader feel the promise of the chapter, visualize the product, or remember the idea before learning details. It should contain more scene, object, UI mockup, gesture, and visual storytelling than text. Keep it book-safe: no garbled text, no unreadable UI, no crowded details, no extreme cinematic/anime/photorealistic style drift. If labels are needed, overlay them afterward with reliable Traditional Chinese text rendering.

Print-Safe Screenshot Direction

For Actual screenshot figures, prioritize print readability and a clean teaching context over faithfully capturing the author's current working desktop.

Use /Users/eric861129/HLR as the default VS Code sample project for book screenshots. This is the clean Habit Life RPG example project. Do not use the publishing/workbook repository for instructional screenshots unless the image is explicitly about publishing files, image production, or book operations.

For VS Code / IDE screenshots:

• Use a light IDE theme, preferably Default Light Modern, for print clarity.
• Use a screenshot-specific workspace/profile when possible, so the author's everyday VS Code layout is not permanently changed.
• Increase editor text enough for print, usually around 15-17px on screen before capture.
• Hide or avoid noisy files and folders in Explorer: node_modules, dist, .pytest_cache, __pycache__, *.db, egg-info, raw screenshots, generated build output, and unrelated private files.
• Show Explorer only when the figure teaches file placement or project structure. For code/preview comparisons, hide Explorer and focus on the editor panes.
• Open only the files needed for the figure. Close unrelated tabs, sidebars, notifications, chat panes, and extension panels.
• Crop to the action and verification point. Avoid macOS menu bar, bookmarks, personal account areas, and unrelated window chrome unless they are needed for the lesson.
• Re-check screenshots against the exact image ID purpose before marking ready. A technically correct screenshot can still be rejected if the book page would feel noisy, dark, or hard to read.

Required Context

Before producing or reviewing an image, read the relevant project files:

• docs/publishing/image-production-checklist.md to find the requested image ID, type, and completion state.
• docs/publishing/image-shot-list.md to confirm priority, core purpose, source manuscript, and status.
• docs/publishing/image-production-guide.md for project naming, screenshot safety, and delivery conventions.
• The source manuscript listed for the image ID when the figure depends on chapter content.

For detailed print, composition, text, and image-type rules, read references/print-image-spec.md.

Workflow

1. Resolve the image ID and source context. If no ID is supplied, ask for one or pick the most relevant pending item only when the user clearly asked you to continue production work.
2. Classify the image type and production path from the checklist, shot list, and source manuscript: 高畫面感插圖, Style A+ diagram, actual screenshot, Mermaid/technical diagram, or QR/external asset.
3. Choose the ratio from the checklist's 建議比例 and the figure purpose. 3:2 is not the default for all images; use 1:1, 4:3, 16:9, 4:5, 3:4, or 15:21 when the page job fits better. Keep critical content away from page edges and any center fold/gutter.
4. Choose the visual style from the image type. Use 高畫面感插圖 for big metaphors and product vision, Style A+ for precise diagrams/comparisons/processes, actual screenshots for operations, and Mermaid for editable technical flows. Keep images close enough that the book feels like one visual series, not a bundle of unrelated images.
5. Plan text handling. Use Traditional Chinese for labels, but avoid asking raster image generation models to render substantial Chinese text. Prefer generating clean art without embedded text, then overlay labels with HTML/CSS, Figma, Illustrator, PowerPoint, or another reliable text renderer. Keep figure text minimal and let icons, shapes, arrows, and layout do more of the explanation.
6. Produce the generation brief, prompt, negative prompt, post-processing plan, target filename, and QA checklist.
7. If a finished image already exists, review it against the spec before marking it ready.

Output Format

When preparing an image brief, return these sections:

• Image ID
• Purpose
• Image type
• Production path
• Book placement
• Dimensions and safe area
• Composition
• Style direction
• Text plan
• Generation prompt
• Negative prompt
• Post-processing
• Filename
• QA checklist

Use concise Traditional Chinese for user-facing labels, prompts, and QA notes unless the user asks otherwise. Keep English technical tokens unchanged when they are product names, filenames, API routes, table names, or code identifiers.

Hard Rules

• Never accept garbled, pseudo-Chinese, simplified Chinese, warped labels, or unreadable Traditional Chinese in final book images.
• Never place essential text, faces, arrows, UI controls, or key diagram nodes near a center fold or inside the inner gutter danger area.
• Never accept a figure that is only readable at screen size. General figures must remain readable when placed about 13cm wide in the book; rare horizontal full-page exceptions must fit within 13cm x 18cm.
• Never make a figure so wide or dense that it only works as a full-bleed web banner; the book page is the primary surface.
• Never treat a non-screenshot figure as only decorated text. If the core idea can be shown with icons, objects, paths, gates, or before/after structure, show it visually and reduce words.
• Never force a 高畫面感插圖, concept illustration, or product vision scene into Style A+ when the manuscript needs emotional memory, atmosphere, or a concrete product vision.
• Never let scenario illustrations swing wildly between unrelated styles such as photorealistic, anime, 3D mascot, dark sci-fi poster, and flat corporate vector unless the user explicitly requests a deliberate one-off exception.
• Never expose API keys, tokens, real email addresses, payment details, tenant/subscription IDs, connection strings, private repo URLs, or private local paths in screenshots.
• Never use dark IDE/tool screenshots for normal instructional figures unless the manuscript specifically teaches dark mode. Prefer light backgrounds for print.
• Never leave a UI action unmarked in the figure when the manuscript tells the reader to click or inspect a specific button, menu, tab, or field. Add a clear callout, frame, or arrow.
• Never default every figure to 3:2. Ratio variety is part of the book design: use square, short-vertical, full-page portrait, 4:3, 3:2, and 16:9 intentionally.
• Preserve editable sources for diagrams and text-heavy figures whenever practical.
• Treat QR Codes as formal image assets: every QR Code needs an original image file, a QR X-X production identifier, and a print-size scan check.

Project Defaults

• Book trim size: 17cm wide x 23cm high.
• Publisher image placement limit: general figures should be designed to work at 13cm wide; rare horizontal full-page exceptions must fit within 13cm x 18cm.
• Safe single-page working area: about 15cm wide x 21cm high.
• Full-page portrait ratio: 15:21 / 5:7, target 3000x4200px or larger; use only for explicitly requested full-page figures.
• Square single-page ratio: 1:1, target 3000x3000px, designed around a 15cm x 15cm page plate.
• Preferred non-full-page ratios: 4:3, 3:2, 16:9, 1:1, 4:5, or 3:4, chosen per image ID.
• Discouraged ratio: any long vertical or ultra-wide figure that becomes unreadable inside a single page, unless it is an explicitly requested 15:21 full-page figure.
• Minimum source quality: 300 dpi at final print size; prefer 450-600 dpi for line art and text-heavy figures.
• Primary language: Traditional Chinese.
• Tone: clear, bright, color-rich professional technical-book visuals. Use Style A+ for diagrams and direct editorial illustration for concept/product-vision scenes. Favor restrained blueprint/magic metaphors, generous whitespace, strong iconography or object storytelling, readable paths, and consistent book-wide illustration quality.