article-illustration-planner

Article Illustration Planner (Article Illustration Planning and Illustration Backfilling)

Deeply analyze the article's structure and cognitive flow, precisely plan a high-quality illustration scheme, and deliver the final illustrated document with completed layout and text trimming.

---

Core Design Philosophy: Genre Adaptation and Cognitive Load Reduction

Before planning any illustration, the Skill must first determine the article's genre and automatically adapt to two completely different working modes:

Mode A: Knowledge / Practical / Tutorial / Hands-On Articles (Knowledge & Explanatory Articles)

Core Mission: Cognitive Load Reduction and High-Fidelity Visual Replacement (High-Fidelity Text-to-Diagram Replacement)

The biggest pain point of knowledge-based articles is dry and lengthy mechanism deduction, hierarchical funnels, formula algorithms, and multi-branch logic. The core mission of illustrations is **"to assist understanding and reduce the reader's burden"**. To this end, Mode A strictly establishes the following four red-line iron rules and execution standards:

1. The Four Absolute Redlines (Hard Constraints & Redlines)

• Redline 1: Miniature Photography and Physical Desk Staging Are Strictly Prohibited (No Miniature / Physical Desk Props)：
  • It is strictly prohibited to include in prompts any physical staged or miniature scenes such as "miniature photography, realistic wooden desk, distressed whiteboard/blackboard, coffee cup, pen, magnifying glass, books, piles of everyday clutter, large-aperture background blur (Bokeh)"!
  • Physical environments and depth-of-field blur consume 30%~40% of the image area, severely compressing the visible size of core charts and creating serious visual noise on mobile devices;
  • The image must be 100% focused on the chart itself, using clean negative space or a light solid-color background, with a full and clear information architecture.
• Redline 2: Irrelevant Narrative Characters and Fairy-Tale Anthropomorphic Elements Are Strictly Prohibited (No Irrelevant Narrative Characters / Fairy Dolls)：
  • It is strictly prohibited to introduce decorative narrative elements such as fairy-tale elves, magic wands, guiding dolls, little guide characters, or unrelated flowers, plants, and animals;
  • The protagonists of knowledge illustrations aregeometric flows, information cards, core data, step labels, and structural logic, absolutely not "toy figures posing in front of a blackboard."
• Redline 3: Core Knowledge Castration Is Strictly Prohibited (Zero Knowledge Loss / Non-Destructive Pruning)：
  • "Text replacement" is essentially about reducing reading resistance, absolutely not "deleting the core substance and replacing it with a vague large image"！
  • [Hardcore Assets That Are Absolutely Prohibited to Delete]：
    1. Mathematical formulas and calculation weights (such as the Xiaohongshu CES scoring formula: likes×1 + saves×1 + comments×4 + shares×4 + follows×8);
    2. Explicit quantitative thresholds and judgment criteria (such as CTR > 8%, impressions < 200, completion rate < 15%);
    3. Key business lines and dialogue templates (such as specific line examples for the four-part account launch script);
    4. Legal/compliance pitfall-avoidance redlines and core platform taboos.
  • [High-Fidelity Replacement Preservation Mechanism]：
    • The chart itselfmust directly draw the core formulas, weight numbers, and judgment thresholds into the image as key visual labels；
    • Refined Reading Guide for the Replaced Main TextThe core conclusions and numerical values must continue to be retained (presented in the form of highlighted formula blocks or a minimalist checklist), ensuring that readers, whether they only look at the images or only read the text, will not encounter knowledge gaps and information devaluation!
• Red Line 4: Empty text relocation is strictly prohibited; true graphical metaphors (True Visual Metaphors) must be used：
  • The reason illustrations can reduce cognitive load is that they replace textual derivation with space and geometric shapes:
    • Hierarchical transition/pool promotion $\to$ Ascending staircase / pyramid(e.g., IG-010 stage progression,IG-001 pyramid);
    • Data loss/filtering conversion $\to$ Progressively narrowing funnel(e.g., IG-013 funnel type);
    • Weight comparison/option trade-offs $\to$ Balance scale / lever / two-column dissection(e.g., IG-004 two-column comparison);
    • Interface structure/system breakdown $\to$ Central main image annotation(e.g., IG-002 central annotation);
    • Multi-branch troubleshooting/SOP $\to$ Decision flow tree(e.g., IG-008 process steps);
    • Psychological anchoring comparison $\to$ Three-column high-low trapezoid / visual gaze focus guidance；
  • It is strictly prohibited to just draw a few boxes and copy the text once!

2. Native Collaboration with 165 Layout Pattern Library (Layout Library Integration)

Knowledge-based illustrationsMandatory precise matching of pattern IDs from the project layout pattern library (especially the 35 types of IG-* infographics and 21 types of SC-* social media cards)! Directly inherit the native composition instructions of the layout pattern (solid color background, comfortable white space, geometric structure), then inject the business theme, step labels, hand-drawn style (#001–#331, prioritizing pure, clean line art/flat/vector styles such as #001 极简白描、#054 包豪斯几何、#216 扁平矢量设计、#018 极简冷面幽默) and classic theme colors (C-01–C-36).

3. Bidirectional Spacing Constraints and Full-Document Chapter Audit (Balanced Spacing & Global Chapter Audit)

• Minimum Physical Separation Distance Constraint (Minimum Separation Distance): Between two adjacent illustrations,the main text interval must strictly not be less than 400–500 characters, and it must span at least one complete second-level/third-level heading. It is strictly prohibited to insert illustrations continuously within the same subsection!
• Maximum Physical Separation Distance Red Line (Maximum Separation Distance): For long-form tutorials (>2500 characters), the main textmust strictly not be completely without illustrations for more than 1000–1200 consecutive characters (or spanning more than 2 major chapters)! Eliminate "dense at the front and empty at the back" or "a full two-chapter gap in the middle";
• Full-Document Grid-Based Scanning Coverage: A full-document inspection of major chapters must be conducted to ensure that key modules such as [Underlying Principles], [Basic Infrastructure/Preparation], [Core Positioning/Benchmarking], [Industrialized SOP], [Data Diagnosis/Troubleshooting], and [Commercial Monetization/Closed Loop] all have balanced visual anchors;
  • 20004000 characters: recommended 34 illustrations;
  • 40008000 words: Recommendation 57 images.

---

Mode B: Non-Knowledge Articles (Lyrical Prose, Narrative Stories, Essays, Novels, etc.) (Non-Knowledge Articles)

Core Mission: Atmospheric Resonance & Visual Pauses (Atmospheric Resonance & Visual Pauses)

1. Preserve the original text without deletion or modification (Preserve Mode): Do not force replacement or modification of the original text; fully preserve the author's narrative and writing style;
2. The style may maintain existing artistic exploration: Visual styles such as narrative macro, poetic metaphor, and hand-drawn illustration may be freely used, with the focus on heightening emotion, creating atmosphere, and providing a sense of visual breathing in the reading rhythm;
3. Placement points follow natural narrative transitions: Choose to insert at points of emotional shift, scene change, or temporal/spatial leap.

---

High Cognitive Load Paragraph Scanning Library for Knowledge Articles (7 Major Image-for-Text Scenarios and Layout Image Type Matching)

When processing knowledge articles, the Skill must proactively search the article and anchor the following 7 types of paragraphs with the greatest "image-for-text" value, and bind recommended layout image types:

High Cognitive Load Type	Typical Features of the Original Text	Bound Recommended Layout Image Type (Layout ID)	Core Fidelity Elements (Must Be Preserved)	Diagram Visual Concept and Load-Reduction Mechanism
1. Algorithms and Weighted Mathematical Models	Includes weighted formulas, multiplier weights, multi-metric composite scoring	IG-001 (Pyramid Infographic) or IG-006 (Multi-Layer Comparison)	Formula body, weight numbers for each item (e.g., ×1, ×4, ×8)	Illustrate the dimension weight balance or composite pyramid; the text retains the formula block, and the diagram presents the weighting hierarchy.
2. Tiered Leaps and Level Advancement	Tier pool advancement, traffic pool levels, long ASCII blocks	IG-010 (Stage-Level Progression Infographic)	Exposure ranges for each tier (e.g., 200-500, 1000-5000), core indicators for advancement assessment	Stepwise progression diagram, arrows pointing upward,directly and completely replace the original ASCII blocks。
3. Funnel Conversion and Layer-by-Layer Loss	Funnel diagnosis, exposure/click/read/interaction/conversion matrix	IG-013 (Funnel Conversion Infographic)	Quantitative standards for each layer (e.g., CTR>8%, completion rate<15%), corresponding optimization actions	A funnel wide at the top and narrow at the bottom, with breakpoint optimization actions marked on the right, intuitively showing layer-by-layer filtering.
4. Interface Anatomy and System Modules	Homepage five-piece set, hardware configuration network isolation, module anatomy	IG-002 (Center Main Image Annotation Infographic)	Core element checklist, functional positioning, pitfall-avoidance requirements	The center is a prototype/mobile interface outline, with radiating leader lines around it annotating the requirements and risk areas of each module.
5. Positive and Negative Cases and Multi-Option Comparison	Viral hit vs. failure comparison, before-and-after pain point contrast, three major business paths	IG-004 (Two-Column Multi-Row Comparison) or IG-011 (Before-and-After Comparison)	Comparison dimensions, applicable audiences, conversion paths	Clear left/right or two-column layout, with inefficient/failed examples on the left and efficient/benchmark examples on the right, clear at a glance.
6. Industrialized SOP and Execution Pipeline	Multi-step sequence, multi-branch decision troubleshooting, four-segment script	IG-008 (Multi-path process step diagram)	Key dialogue script examples, sequence order, troubleshooting branches	Horizontal/S-shaped flowing step bar with icons, clear nodes, readable arrows.
7. Price anchoring and product matrix	Multi-SKU combinations, pricing psychology, hero product vs. foil product	IG-007 (Label thumbnail card) or SC-004 (Image-text split card)	Price figures (19.9/35.9/59.9), best-selling share (70%), anchoring effect	Three-column side-by-side cards, highlighting the middle hero product, with psychological anchoring effect arrows labeled.

---

Input parameters and configuration specifications

1. Article content or local path: Paste the body text directly, or provide a local path (e.g., D:\path\to\article.md）。
2. Hand-drawn style and color tone：
   • Style code: full library #001–#331；
   • Knowledge article recommendation guidelines: Strictly prohibited to recommend children's fairy-tale style or abstract scribble style; prioritize recommending pure, clean styles with clear line-art blocks and surfaces that facilitate information delivery (e.g., #001 极简白描、#054 包豪斯几何构成、#216 扁平矢量设计、#018 极简冷面幽默 etc.);
   • Theme color code: full library C-01–C-36, prioritize steady, professional tones (e.g., C-01 经典蓝、C-10 鼠尾草绿、C-31 佩恩灰、C-05 暖赭石、C-22 普鲁士蓝）；
   • Dynamic assessment: If unspecified, the AI dynamically recommends a single style and single color tone that runs through the entire piece based on the article context (strictly prohibited to have a chaotic mix of styles throughout);.
3. Aspect Ratio：
   • Default 4:3(Standard for long-form horizontal reading flow); mobile social cards support 3:4 or 16:9。
4. Whitespace & Background：
   • Knowledge diagrams require a strictly pure background and comfortable whitespace; cluttered backgrounds are strictly prohibited.

---

Standard planning solution output template (Output Format)

First, issue overall design decisions:

• Article type determination: Knowledge-based / Tutorial-based / Lyrical prose / Narrative story
• Applicable work mode: Mode A (cognitive load reduction and high-fidelity diagrams replacing lengthy text) / Mode B (atmosphere enhancement and preservation mode)
• Overall visual specifications for the entire piece: Layout diagram type tone (IG/SC), hand-drawn style code and name, theme color code and name, unified aspect ratio (4:3）
• Full-piece chapter coverage and spacing audit: List the chapters of the entire piece, mark the chapters covered by each illustration, and confirm that both the minimum spacing (>400 characters) and maximum spacing (<1200 characters) are compliant
• Recommended total number of illustrations: Usually 3–6 high-value points (balanced throughout the entire piece, no gaps, no stacking)

Then, issue the illustration plan for each image in reading order:

### Illustration N

**插入/替换位置 / Anchor:**
`[引用紧邻的小标题或上下文关键句（10-20字）]`

**配图定位 / Purpose:**
解释该图在阅读流中的认知价值与承接逻辑。

**排版图型 / Layout ID:**
`IG-013` (漏斗转化型信息图) / `IG-010` (阶段进阶型信息图) / `IG-002` (中心标注信息图) 等

**图片类型 / Image Type:**
process diagram / conceptual diagram / comparison diagram / infographic

**核心保真要素 / Core Knowledge to Preserve:**
[明确列出本图必须保留的数学公式、量化指标数字（如>8%）、关键话术台词或业务红线]

**图文关系 / Relationship to Text:**
replaces complex text (文字替换与认知降负) / supplements text (内容补充) / visual-pause (视觉停顿)

**（若为文字替换）被图片替代的原繁冗文本 / Text to Replace:**
> [明确摘录原文中将被该图替代并删除的复杂文字推导、长篇ASCII图表或嵌套列表片段]

**（若为文字替换）精炼后的轻量导读 / Simplified Replacement Text:**
[文字替换后在 Markdown 正文中保留的极简核心要点/行动清单，必须完整保留核心数值与公式结论，通常为 1-3 句话]

**图形隐喻设计 / Visual Metaphor:**
[明确说明用何种几何隐喻降低认知负荷，如上宽下窄漏斗、上升阶梯、对比天平]

**画风与配色 / Style & Color:**
`#{style_number} · {style_name}` + `{color_id} · {color_name}`

**生图提示词 (Prompt):**
- **中文提示词**:
```text
小红书信息图。排版：{layout_prompt}。纯净背景，大面积舒适留白，严禁桌椅、文具、咖啡杯等任何物理杂物摆设，严禁童话/小精灵角色，图表主体占满画幅90%以上。核心内容：{visual_metaphor_and_labels}。画风：{style_prompt}。色彩：{color_prompt}。画幅比例 4:3。

• English Prompt:

Xiaohongshu infographic. Layout: {layout_prompt_en}. Clean background with generous negative space, strictly NO physical desks, stationery, coffee cups or props, strictly NO fairy/doll characters, chart fills over 90% of frame. Core content: {visual_metaphor_and_labels_en}. Style: {style_prompt_en}. Color palette: {color_prompt_en}. Aspect ratio: 4:3.

---

## 交付与后续操作引导 (CTA)

规划出具后，必须出具标准化双轨交付引导：

```markdown
---

💡 **插图方案已规划完成！接下来您可以选择以下两种交付方式：**

1. **【方式 A · 自主生图回填】**：
   复制上方提示词，前往您常用的生图工具出图。生成完成后，将图片文件直接拖入对话或发送本地图片路径，我将自动帮您精简替换原文繁冗文字，输出完整的图文定稿！

2. **【方式 B · 全自动生图插入】**：
   直接对我说 **“全自动生图”** 或 **“帮我生成所有配图并插入”**，我将全自动调用生图工具批量绘制配图，自动执行文字替换与排版回填，一键输出排版完毕的 Markdown 定稿文件！

---

Dual-Track Assembly

When the user triggers backfilling:

1. Automatically organize resources: Save the images with standardized names as <article_dir>/images/illus_01.webp(or .png）；
2. Perform image-text assembly and high-fidelity text replacement：
   • If a knowledge article defines Text to Replace: Automatically trim and replace the corresponding complex and lengthy original textTrim and replacewith illustration Markdown,*▲ 图N：说明*and[精炼导读/公式清单]; Fill in image addresses according to the actual files for this issueimages/illus_编号.webp, not static files bundled with the package;
   • Strictly executeFidelity check: Ensure the refined introduction retains the original core formulas and data;
   • If no text replacement is defined: the specification inserts the image after the specified anchor point;
3. Output final version: Generate [article_name]_illustrated.md。

It can also be completed by calling the built-in deterministic automation script:

python -X utf8 skills/article-illustration-planner/scripts/insert_illustrations.py \
  --article "D:\path\to\article.md" \
  --manifest "D:\path\to\manifest.json" \
  --output "D:\path\to\article_illustrated.md"