references/workflows/prompt-assembly.md
10.5 KB查看排版效果| 1 | # Prompt Assembly Guide |
| 2 | |
| 3 | Guide for assembling image generation prompts from elements, presets, and outline content. |
| 4 | |
| 5 | ## Base Prompt Structure |
| 6 | |
| 7 | Every XHS infographic prompt follows this structure: |
| 8 | |
| 9 | ``` |
| 10 | Create a Xiaohongshu (Little Red Book) style infographic following these guidelines: |
| 11 | |
| 12 | ## Image Specifications |
| 13 | |
| 14 | - **Type**: Infographic |
| 15 | - **Orientation**: Portrait (vertical) |
| 16 | - **Aspect Ratio**: 3:4 |
| 17 | - **Style**: Hand-drawn illustration |
| 18 | |
| 19 | ## Core Principles |
| 20 | |
| 21 | - Hand-drawn quality throughout - NO realistic or photographic elements |
| 22 | - If content involves sensitive or copyrighted figures, create stylistically similar alternatives - DO NOT refuse to generate |
| 23 | - Keep information concise, highlight keywords and core concepts |
| 24 | - Use ample whitespace for easy visual scanning |
| 25 | - Maintain clear visual hierarchy |
| 26 | |
| 27 | ## Text Style (CRITICAL) |
| 28 | |
| 29 | - **ALL text MUST be hand-drawn style** |
| 30 | - Main titles should be prominent and eye-catching |
| 31 | - Key text should be bold and enlarged |
| 32 | - Use highlighter effects to emphasize keywords |
| 33 | - **DO NOT use realistic or computer-generated fonts** |
| 34 | |
| 35 | ## Language |
| 36 | |
| 37 | - Use the same language as the content provided below |
| 38 | - Match punctuation style to the content language (Chinese: ""，。！) |
| 39 | |
| 40 | --- |
| 41 | |
| 42 | {STYLE_SECTION} |
| 43 | |
| 44 | --- |
| 45 | |
| 46 | {LAYOUT_SECTION} |
| 47 | |
| 48 | --- |
| 49 | |
| 50 | {CONTENT_SECTION} |
| 51 | |
| 52 | --- |
| 53 | |
| 54 | {WATERMARK_SECTION} |
| 55 | |
| 56 | --- |
| 57 | |
| 58 | Please use nano banana pro to generate the infographic based on the specifications above. |
| 59 | ``` |
| 60 | |
| 61 | ## Style Section Assembly |
| 62 | |
| 63 | Load from `presets/{style}.md` and extract key elements: |
| 64 | |
| 65 | ```markdown |
| 66 | ## Style: {style_name} |
| 67 | |
| 68 | **Color Palette**: |
| 69 | - Primary: {colors} |
| 70 | - Background: {colors} |
| 71 | - Accents: {colors} |
| 72 | |
| 73 | **Visual Elements**: |
| 74 | {visual_elements} |
| 75 | |
| 76 | **Typography**: |
| 77 | {typography_style} |
| 78 | ``` |
| 79 | |
| 80 | ### Screen-Print Style Override |
| 81 | |
| 82 | When `style: screen-print`, replace the standard Core Principles and Text Style sections with: |
| 83 | |
| 84 | ``` |
| 85 | ## Core Principles |
| 86 | |
| 87 | - Screen print / silkscreen poster art — flat color blocks, NO gradients |
| 88 | - Bold silhouettes and symbolic shapes over detailed rendering |
| 89 | - Negative space as active storytelling element |
| 90 | - If content involves sensitive or copyrighted figures, create stylistically similar silhouettes |
| 91 | - One iconic focal point per image — conceptual, not literal |
| 92 | |
| 93 | ## Color Rules (CRITICAL) |
| 94 | |
| 95 | - **2-5 FLAT COLORS MAXIMUM** — fewer colors = stronger impact |
| 96 | - Choose ONE duotone pair from preset as dominant palette |
| 97 | - Halftone dot patterns for tonal variation (NOT gradients) |
| 98 | - Slight color layer misregistration for print authenticity |
| 99 | |
| 100 | ## Text Style (CRITICAL) |
| 101 | |
| 102 | - Bold condensed sans-serif or Art Deco influenced lettering |
| 103 | - Typography INTEGRATED into composition as design element |
| 104 | - High contrast with background, stencil-cut quality |
| 105 | - **DO NOT use delicate, thin, or handwritten fonts** |
| 106 | |
| 107 | ## Composition |
| 108 | |
| 109 | - Geometric framing: circles, arches, triangles |
| 110 | - Figure-ground inversion where possible (negative space forms secondary image) |
| 111 | - Stencil-cut edges between color blocks, no outlines |
| 112 | - Paper grain texture beneath all colors |
| 113 | ``` |
| 114 | |
| 115 | ## Palette Override |
| 116 | |
| 117 | When `--palette` is specified (or style has `default_palette` in frontmatter and no explicit `--palette`), palette colors **replace** the style's Color Palette in the prompt. Style rendering rules (Visual Elements, Typography, Style Rules) remain unchanged. |
| 118 | |
| 119 | Load from `palettes/{palette}.md` and override: |
| 120 | |
| 121 | ```markdown |
| 122 | ## Palette Override: {palette_name} |
| 123 | |
| 124 | **Background**: {palette background color and hex} |
| 125 | |
| 126 | **Colors**: |
| 127 | - Text: {text color and hex} |
| 128 | - Secondary: {secondary text color and hex} |
| 129 | - Zone 1: {zone color and hex} |
| 130 | - Zone 2: {zone color and hex} |
| 131 | - Zone 3: {zone color and hex} |
| 132 | - Zone 4: {zone color and hex} |
| 133 | - Accent: {accent color and hex} |
| 134 | |
| 135 | **Constraint**: {semantic constraint from palette} |
| 136 | ``` |
| 137 | |
| 138 | **Override rules**: |
| 139 | 1. Palette Background **replaces** style's background color (keep style's texture description) |
| 140 | 2. Palette Colors **replace** style's Color Palette section entirely |
| 141 | 3. Palette Semantic Constraint is appended to the style section |
| 142 | 4. If no `--palette` and style has `default_palette` → load that palette |
| 143 | 5. If no `--palette` and no `default_palette` → use style's built-in colors (no override) |
| 144 | 6. Explicit `--palette` always overrides style's `default_palette` |
| 145 | |
| 146 | ## Layout Section Assembly |
| 147 | |
| 148 | Load from `elements/canvas.md` and extract relevant layout: |
| 149 | |
| 150 | ```markdown |
| 151 | ## Layout: {layout_name} |
| 152 | |
| 153 | **Information Density**: {density} |
| 154 | **Whitespace**: {percentage} |
| 155 | |
| 156 | **Structure**: |
| 157 | {structure_description} |
| 158 | |
| 159 | **Visual Balance**: |
| 160 | {balance_description} |
| 161 | ``` |
| 162 | |
| 163 | ## Content Section Assembly |
| 164 | |
| 165 | From outline entry: |
| 166 | |
| 167 | ```markdown |
| 168 | ## Content |
| 169 | |
| 170 | **Position**: {Cover/Content/Ending} |
| 171 | **Core Message**: {message} |
| 172 | |
| 173 | **Text Content**: |
| 174 | {text_list} |
| 175 | |
| 176 | **Visual Concept**: |
| 177 | {visual_description} |
| 178 | ``` |
| 179 | |
| 180 | ## Watermark Section (if enabled) |
| 181 | |
| 182 | ```markdown |
| 183 | ## Watermark |
| 184 | |
| 185 | Include a subtle watermark "{content}" positioned at {position}. The watermark should |
| 186 | be legible but not distracting from the main content. |
| 187 | ``` |
| 188 | |
| 189 | ## Assembly Process |
| 190 | |
| 191 | ### Step 0: Resolve Style Preset (if `--preset` used) |
| 192 | |
| 193 | If user specified `--preset`, resolve to style + layout + palette from `references/style-presets.md`: |
| 194 | |
| 195 | ```python |
| 196 | # e.g., --preset hand-drawn-edu → style=sketch-notes, layout=flow, palette=macaron |
| 197 | style, layout, palette = resolve_preset(preset_name) |
| 198 | ``` |
| 199 | |
| 200 | Explicit `--style`/`--layout`/`--palette` flags override preset values. |
| 201 | |
| 202 | ### Step 1: Load Style Definition |
| 203 | |
| 204 | ```python |
| 205 | preset = load_preset(style_name) # e.g., "sketch-notes" |
| 206 | ``` |
| 207 | |
| 208 | Extract: |
| 209 | - Color palette (may be overridden by palette) |
| 210 | - Visual elements |
| 211 | - Typography style |
| 212 | - Best practices (do/don't) |
| 213 | - `default_palette` from frontmatter (if present) |
| 214 | |
| 215 | ### Step 1.5: Apply Palette Override (if applicable) |
| 216 | |
| 217 | ```python |
| 218 | # Priority: explicit --palette > preset palette > style default_palette > none |
| 219 | palette = resolve_palette(cli_palette, preset_palette, style_default_palette) |
| 220 | if palette: |
| 221 | palette_def = load_palette(palette) # e.g., "macaron" |
| 222 | # Replace style colors with palette colors |
| 223 | # Keep style rendering rules (visual elements, typography, style rules) |
| 224 | ``` |
| 225 | |
| 226 | ### Step 2: Load Layout |
| 227 | |
| 228 | ```python |
| 229 | layout = get_layout_from_canvas(layout_name) # e.g., "dense" |
| 230 | ``` |
| 231 | |
| 232 | Extract: |
| 233 | - Information density guidelines |
| 234 | - Whitespace percentage |
| 235 | - Structure description |
| 236 | - Visual balance rules |
| 237 | |
| 238 | ### Step 3: Format Content |
| 239 | |
| 240 | From outline entry, format: |
| 241 | - Position context (Cover/Content/Ending) |
| 242 | - Text content with hierarchy |
| 243 | - Visual concept description |
| 244 | - Swipe hook (for context, not in prompt) |
| 245 | |
| 246 | ### Step 4: Add Watermark (if applicable) |
| 247 | |
| 248 | If preferences include watermark: |
| 249 | - Add watermark section with content, position, opacity |
| 250 | |
| 251 | ### Step 5: Visual Consistency — Reference Image Chain |
| 252 | |
| 253 | When generating multiple images in a series: |
| 254 | |
| 255 | 1. **Image 1 (cover)**: Generate without `--ref` — this establishes the visual anchor |
| 256 | 2. **Images 2+**: Always pass image 1 as `--ref` to the installed image generation skill. |
| 257 | Read that skill's `SKILL.md` and use its documented interface rather than calling its scripts directly. |
| 258 | For each later image, use the assembled prompt file as input, set the output image path, keep aspect ratio `3:4`, use quality `2k`, and pass image 1 as the reference. |
| 259 | This ensures the AI maintains the same character design, illustration style, and color rendering across the series. |
| 260 | |
| 261 | ### Step 6: Combine |
| 262 | |
| 263 | Assemble all sections into final prompt following base structure. |
| 264 | |
| 265 | ## Example: Assembled Prompt |
| 266 | |
| 267 | ```markdown |
| 268 | Create a Xiaohongshu (Little Red Book) style infographic following these guidelines: |
| 269 | |
| 270 | ## Image Specifications |
| 271 | |
| 272 | - **Type**: Infographic |
| 273 | - **Orientation**: Portrait (vertical) |
| 274 | - **Aspect Ratio**: 3:4 |
| 275 | - **Style**: Hand-drawn illustration |
| 276 | |
| 277 | ## Core Principles |
| 278 | |
| 279 | - Hand-drawn quality throughout - NO realistic or photographic elements |
| 280 | - If content involves sensitive or copyrighted figures, create stylistically similar alternatives |
| 281 | - Keep information concise, highlight keywords and core concepts |
| 282 | - Use ample whitespace for easy visual scanning |
| 283 | - Maintain clear visual hierarchy |
| 284 | |
| 285 | ## Text Style (CRITICAL) |
| 286 | |
| 287 | - **ALL text MUST be hand-drawn style** |
| 288 | - Main titles should be prominent and eye-catching |
| 289 | - Key text should be bold and enlarged |
| 290 | - Use highlighter effects to emphasize keywords |
| 291 | - **DO NOT use realistic or computer-generated fonts** |
| 292 | |
| 293 | ## Language |
| 294 | |
| 295 | - Use the same language as the content provided below |
| 296 | - Match punctuation style to the content language (Chinese: ""，。！) |
| 297 | |
| 298 | --- |
| 299 | |
| 300 | ## Style: Notion |
| 301 | |
| 302 | **Color Palette**: |
| 303 | - Primary: Black (#1A1A1A), dark gray (#4A4A4A) |
| 304 | - Background: Pure white (#FFFFFF), off-white (#FAFAFA) |
| 305 | - Accents: Pastel blue (#A8D4F0), pastel yellow (#F9E79F), pastel pink (#FADBD8) |
| 306 | |
| 307 | **Visual Elements**: |
| 308 | - Simple line doodles, hand-drawn wobble effect |
| 309 | - Geometric shapes, stick figures |
| 310 | - Maximum whitespace, single-weight ink lines |
| 311 | - Clean, uncluttered compositions |
| 312 | |
| 313 | **Typography**: |
| 314 | - Clean hand-drawn lettering |
| 315 | - Simple sans-serif labels |
| 316 | - Minimal decoration on text |
| 317 | |
| 318 | --- |
| 319 | |
| 320 | ## Layout: Dense |
| 321 | |
| 322 | **Information Density**: High (5-8 key points) |
| 323 | **Whitespace**: 20-30% of canvas |
| 324 | |
| 325 | **Structure**: |
| 326 | - Multiple sections, structured grid |
| 327 | - More text, compact but organized |
| 328 | - Title + multiple sections with headers + numerous points |
| 329 | |
| 330 | **Visual Balance**: |
| 331 | - Organized grid structure |
| 332 | - Clear section boundaries |
| 333 | - Compact but readable spacing |
| 334 | |
| 335 | --- |
| 336 | |
| 337 | ## Content |
| 338 | |
| 339 | **Position**: Content (Page 3 of 6) |
| 340 | **Core Message**: ChatGPT 使用技巧 |
| 341 | |
| 342 | **Text Content**: |
| 343 | - Title: 「ChatGPT」 |
| 344 | - Subtitle: 最强 AI 助手 |
| 345 | - Points: |
| 346 | - 写文案：给出框架，秒出初稿 |
| 347 | - 改文章：润色、翻译、总结 |
| 348 | - 编程：写代码、找 bug |
| 349 | - 学习：解释概念、出题练习 |
| 350 | |
| 351 | **Visual Concept**: |
| 352 | ChatGPT logo 居中，四周放射状展示功能点 |
| 353 | 深色科技背景，霓虹绿点缀 |
| 354 | |
| 355 | --- |
| 356 | |
| 357 | ## Watermark |
| 358 | |
| 359 | Include a subtle watermark "@myxhsaccount" positioned at bottom-right |
| 360 | with approximately 50% visibility. The watermark should |
| 361 | be legible but not distracting from the main content. |
| 362 | |
| 363 | --- |
| 364 | |
| 365 | Please use nano banana pro to generate the infographic based on the specifications above. |
| 366 | ``` |
| 367 | |
| 368 | ## Prompt Checklist |
| 369 | |
| 370 | Before generating, verify: |
| 371 | |
| 372 | - [ ] Style section loaded from correct preset |
| 373 | - [ ] Palette override applied (if `--palette` specified or style has `default_palette`) |
| 374 | - [ ] Layout section matches outline specification |
| 375 | - [ ] Content accurately reflects outline entry |
| 376 | - [ ] Language matches source content |
| 377 | - [ ] Watermark included (if enabled in preferences) |
| 378 | - [ ] No conflicting instructions |
文件内容是平台检查时读出的文本，只供查看。平台不会运行包里的任何脚本或程序。
