diff --git a/.claude-plugin/marketplace.json b/.claude-plugin/marketplace.json
index 719a7bf..f53bb34 100644
--- a/.claude-plugin/marketplace.json
+++ b/.claude-plugin/marketplace.json
@@ -6,7 +6,7 @@
   },
   "metadata": {
     "description": "Skills shared by Baoyu for improving daily work efficiency",
-    "version": "1.18.3"
+    "version": "1.19.0"
   },
   "plugins": [
     {
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 436f379..286bd5e 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -2,6 +2,21 @@
 
 English | [中文](./CHANGELOG.zh.md)
 
+## 1.19.0 - 2026-01-24
+
+### Features
+- `baoyu-comic`: adds partial workflow options—`--storyboard-only`, `--prompts-only`, `--images-only`, and `--regenerate N` for flexible workflow control.
+- `baoyu-image-gen`: adds `--imageSize` parameter for Google providers (1K/2K/4K), changes default quality to 2k.
+- `baoyu-image-gen`: adds `GEMINI_API_KEY` as alias for `GOOGLE_API_KEY`.
+
+### Refactor
+- `baoyu-comic`: extracts detailed workflow to `references/workflow.md`, reduces SKILL.md by ~400 lines while preserving functionality.
+- `baoyu-comic`: extracts content signal analysis to `references/auto-selection.md` and partial workflow docs to `references/partial-workflows.md`.
+- `baoyu-image-gen`: modularizes code—extracts types to `types.ts`, provider implementations to `providers/google.ts` and `providers/openai.ts`.
+
+### Documentation
+- `baoyu-comic`: improves ohmsha preset documentation with explicit default Doraemon character definitions and visual descriptions.
+
 ## 1.18.3 - 2026-01-23
 
 ### Documentation
diff --git a/CHANGELOG.zh.md b/CHANGELOG.zh.md
index d3a4e11..c5c2c0c 100644
--- a/CHANGELOG.zh.md
+++ b/CHANGELOG.zh.md
@@ -2,6 +2,21 @@
 
 [English](./CHANGELOG.md) | 中文
 
+## 1.19.0 - 2026-01-24
+
+### 新功能
+- `baoyu-comic`：新增部分工作流选项——`--storyboard-only`、`--prompts-only`、`--images-only` 和 `--regenerate N`，实现灵活的工作流控制。
+- `baoyu-image-gen`：新增 `--imageSize` 参数用于 Google 提供商（1K/2K/4K），默认质量改为 2k。
+- `baoyu-image-gen`：新增 `GEMINI_API_KEY` 作为 `GOOGLE_API_KEY` 的别名。
+
+### 重构
+- `baoyu-comic`：将详细工作流提取至 `references/workflow.md`，SKILL.md 减少约 400 行，功能完整保留。
+- `baoyu-comic`：将内容信号分析提取至 `references/auto-selection.md`，部分工作流文档提取至 `references/partial-workflows.md`。
+- `baoyu-image-gen`：代码模块化——类型定义提取至 `types.ts`，provider 实现提取至 `providers/google.ts` 和 `providers/openai.ts`。
+
+### 文档
+- `baoyu-comic`：改进 ohmsha 预设文档，明确默认哆啦A梦角色定义和视觉描述。
+
 ## 1.18.3 - 2026-01-23
 
 ### 文档
diff --git a/skills/baoyu-comic/SKILL.md b/skills/baoyu-comic/SKILL.md
index c7aff99..3b35bef 100644
--- a/skills/baoyu-comic/SKILL.md
+++ b/skills/baoyu-comic/SKILL.md
@@ -27,6 +27,17 @@ Create original knowledge comics with flexible art style × tone combinations.
 | `--aspect` | 3:4 (default, portrait), 4:3 (landscape), 16:9 (widescreen) | Page aspect ratio |
 | `--lang` | auto (default), zh, en, ja, etc. | Output language |
 
+### Partial Workflow Options
+
+| Option | Description |
+|--------|-------------|
+| `--storyboard-only` | Generate storyboard only, skip prompts and images |
+| `--prompts-only` | Generate storyboard + prompts, skip images |
+| `--images-only` | Generate images from existing prompts directory |
+| `--regenerate N` | Regenerate specific page(s) only (e.g., `3` or `2,5,8`) |
+
+Details: [references/partial-workflows.md](references/partial-workflows.md)
+
 ### Art Styles (画风)
 
 | Style | 中文 | Description |
@@ -69,27 +80,25 @@ Presets with special rules beyond art+tone:
 | ink-brush | neutral, dramatic, action, vintage | warm | romantic, energetic |
 | chalk | neutral, warm, energetic | vintage | dramatic, action, romantic |
 
-Art Style × Tone × Layout can be freely combined. Incompatible combinations work but may produce unexpected results.
+Details: [references/auto-selection.md](references/auto-selection.md)
 
 ## Auto Selection
 
 Content signals determine default art + tone + layout (or preset):
 
-| Content Signals | Art Style | Tone | Layout | Preset |
-|-----------------|-----------|------|--------|--------|
-| Tutorial, how-to, beginner | manga | neutral | webtoon | **ohmsha** |
-| Computing, AI, programming | manga | neutral | dense | **ohmsha** |
-| Technical explanation, educational | manga | neutral | webtoon | **ohmsha** |
-| Pre-1950, classical, ancient | realistic | vintage | cinematic | - |
-| Personal story, mentor | ligne-claire | warm | standard | - |
-| Conflict, breakthrough | (inherit) | dramatic | splash | - |
-| Wine, food, business, lifestyle | realistic | neutral | cinematic | - |
-| Martial arts, wuxia, xianxia | ink-brush | action | splash | **wuxia** |
-| Romance, love, school life | manga | romantic | standard | **shoujo** |
-| Biography, balanced | ligne-claire | neutral | mixed | - |
+| Content Signals | Recommended |
+|-----------------|-------------|
+| Tutorial, how-to, programming, educational | **ohmsha** preset |
+| Pre-1950, classical, ancient | realistic + vintage |
+| Personal story, mentor | ligne-claire + warm |
+| Martial arts, wuxia | **wuxia** preset |
+| Romance, school life | **shoujo** preset |
+| Biography, balanced | ligne-claire + neutral |
 
 **When preset is recommended**: Load `references/presets/{preset}.md` and apply all special rules.
 
+Details: [references/auto-selection.md](references/auto-selection.md)
+
 ## Script Directory
 
 **Important**: All scripts are located in the `scripts/` subdirectory of this skill.
@@ -111,409 +120,147 @@ Output directory: `comic/{topic-slug}/`
 - Conflict: append timestamp (e.g., `turing-story-20260118-143052`)
 
 **Contents**:
-- `source-{slug}.{ext}` - Source files
-- `analysis.md` - Content analysis
-- `storyboard.md` - Storyboard with panel breakdown
-- `characters/characters.md` - Character definitions
-- `characters/characters.png` - Character reference sheet
-- `prompts/NN-{cover|page}-[slug].md` - Generation prompts
-- `NN-{cover|page}-[slug].png` - Generated images
-- `{topic-slug}.pdf` - Final merged PDF
+| File | Description |
+|------|-------------|
+| `source-{slug}.{ext}` | Source files |
+| `analysis.md` | Content analysis |
+| `storyboard.md` | Storyboard with panel breakdown |
+| `characters/characters.md` | Character definitions |
+| `characters/characters.png` | Character reference sheet |
+| `prompts/NN-{cover\|page}-[slug].md` | Generation prompts |
+| `NN-{cover\|page}-[slug].png` | Generated images |
+| `{topic-slug}.pdf` | Final merged PDF |
+
+## Language Handling
+
+**Detection Priority**:
+1. `--lang` flag (explicit)
+2. EXTEND.md `language` setting
+3. User's conversation language
+4. Source content language
+
+**Rule**: Use user's input language or saved language preference for ALL interactions:
+- Storyboard outlines and scene descriptions
+- Image generation prompts
+- User selection options and confirmations
+- Progress updates, questions, errors, summaries
+
+Technical terms remain in English.
 
 ## Workflow
 
 ### Progress Checklist
 
-Copy and track progress:
-
 ```
 Comic Progress:
-- [ ] Step 1: Load preferences + Analyze content
-- [ ] Step 2: Confirmation 1 - Style & options ⚠️ REQUIRED
+- [ ] Step 1: Setup & Analyze (1.1 Preferences, 1.2 Analyze, 1.3 Check existing)
+- [ ] Step 2: Confirmation - Style & options ⚠️ REQUIRED
 - [ ] Step 3: Generate storyboard + characters
-- [ ] Step 4: Confirmation 2 - Review outline (if requested)
-- [ ] Step 5: Generate images (sequential)
-- [ ] Step 6: Merge to PDF
-- [ ] Step 7: Completion report
+- [ ] Step 4: Review outline (conditional)
+- [ ] Step 5: Generate prompts
+- [ ] Step 6: Review prompts (conditional)
+- [ ] Step 7: Generate images ⚠️ CHARACTER REF REQUIRED
+  - [ ] 7.1 Generate character sheet FIRST → characters/characters.png
+  - [ ] 7.2 Generate pages WITH --ref characters/characters.png
+- [ ] Step 8: Merge to PDF
+- [ ] Step 9: Completion report
 ```
 
 ### Flow
 
 ```
-Input → Preferences → Analyze → [Confirm 1: Style + Skip?] → Storyboard → [Confirm 2: if needed] → Generate → PDF → Complete
+Input → Preferences → Analyze → [Check Existing?] → [Confirm: Style + Reviews] → Storyboard → [Review?] → Prompts → [Review?] → Images → PDF → Complete
 ```
 
-### Step 1: Setup & Analyze
+### Step Summary
 
-**1.1 Load Preferences (EXTEND.md)**
+| Step | Action | Key Output |
+|------|--------|------------|
+| 1.1 | Load EXTEND.md preferences | Config loaded |
+| 1.2 | Analyze content | `analysis.md` |
+| 1.3 | Check existing directory | Handle conflicts |
+| 2 | Confirm style, focus, audience, reviews | User preferences |
+| 3 | Generate storyboard + characters | `storyboard.md`, `characters/` |
+| 4 | Review outline (if requested) | User approval |
+| 5 | Generate prompts | `prompts/*.md` |
+| 6 | Review prompts (if requested) | User approval |
+| **7.1** | **Generate character sheet FIRST** | `characters/characters.png` |
+| **7.2** | Generate pages **with character ref** | `*.png` files |
+| 8 | Merge to PDF | `{slug}.pdf` |
+| 9 | Completion report | Summary |
 
-Use Bash to check EXTEND.md existence (priority order):
+### Step 7: Image Generation ⚠️ CRITICAL
+
+**Character reference is MANDATORY for visual consistency.**
+
+**7.1 Generate character sheet first**:
+```bash
+# Use Reference Sheet Prompt from characters/characters.md
+npx -y bun ${SKILL_DIR}/../baoyu-image-gen/scripts/main.ts \
+  --promptfiles characters/characters.md \
+  --image characters/characters.png --ar 4:3
+```
+
+**7.2 Generate each page WITH character reference**:
+
+| Skill Capability | Strategy |
+|------------------|----------|
+| Supports `--ref` | Pass `characters/characters.png` with EVERY page |
+| No `--ref` support | Prepend character descriptions to EVERY prompt file |
 
 ```bash
-# Check project-level first
-test -f .baoyu-skills/baoyu-comic/EXTEND.md && echo "project"
-
-# Then user-level (cross-platform: $HOME works on macOS/Linux/WSL)
-test -f "$HOME/.baoyu-skills/baoyu-comic/EXTEND.md" && echo "user"
-```
-
-┌──────────────────────────────────────────────┬───────────────────┐
-│                     Path                     │     Location      │
-├──────────────────────────────────────────────┼───────────────────┤
-│ .baoyu-skills/baoyu-comic/EXTEND.md          │ Project directory │
-├──────────────────────────────────────────────┼───────────────────┤
-│ $HOME/.baoyu-skills/baoyu-comic/EXTEND.md    │ User home         │
-└──────────────────────────────────────────────┴───────────────────┘
-
-**When EXTEND.md Found** → Read, parse, **output summary to user**:
-
-```
-📋 Loaded preferences from [full path]
-├─ Watermark: [enabled/disabled] [content if enabled]
-├─ Art Style: [style name or "auto-select"]
-├─ Tone: [tone name or "auto-select"]
-├─ Layout: [layout or "auto-select"]
-├─ Language: [language or "auto-detect"]
-└─ Character presets: [count] defined
-```
-
-**MUST output this summary** so user knows their current configuration. Do not skip or silently load.
-
-**When EXTEND.md Not Found** → First-time setup:
-
-1. Inform user: "No preferences found. Let's set up your defaults."
-2. Use AskUserQuestion to collect preferences (see `references/config/first-time-setup.md`)
-3. Create EXTEND.md at user-chosen location
-4. Confirm: "✓ Preferences saved to [path]"
-
-**EXTEND.md Supports**: Watermark | Preferred art/tone/layout | Custom style definitions | Character presets | Language preference
-
-Schema: `references/config/preferences-schema.md`
-
-**Important**: Once EXTEND.md exists, watermark, language, and style defaults are NOT asked again in Confirmation 1 or 2. These are session-persistent settings.
-
-**1.2 Analyze Content → `analysis.md`**
-
-Read source content, save it if needed, and perform deep analysis.
-
-**Actions**:
-1. **Save source content** (if not already a file):
-   - If user provides a file path: use as-is
-   - If user pastes content: save to `source.md` in target directory
-2. Read source content
-3. **Deep analysis** following `references/analysis-framework.md`:
-   - Target audience identification
-   - Value proposition for readers
-   - Core themes and narrative potential
-   - Key figures and their story arcs
-4. Detect source language
-5. **Determine language**:
-   - If EXTEND.md has `language` → use it
-   - Else if `--lang` option provided → use it
-   - Else → use detected source language
-6. Determine recommended page count:
-   - Short story: 5-8 pages
-   - Medium complexity: 9-15 pages
-   - Full biography: 16-25 pages
-7. Analyze content signals for art/tone/layout recommendations
-8. **Save to `analysis.md`**
-
-**analysis.md Format**: YAML front matter (title, topic, time_span, source_language, user_language, aspect_ratio, recommended_page_count, recommended_art, recommended_tone) + sections for Target Audience, Value Proposition, Core Themes, Key Figures & Story Arcs, Content Signals, Recommended Approaches. See `references/analysis-framework.md` for full template.
-
-### Step 2: Confirmation 1 - Style & Options ⚠️
-
-**Purpose**: Select visual style + decide whether to review outline before generation. **Do NOT skip.**
-
-**Note**: Watermark and language already configured in EXTEND.md (Step 1).
-
-**Display summary**:
-- Content type + topic identified
-- Key figures extracted
-- Time span detected
-- Recommended page count
-- Language: [from EXTEND.md or detected]
-- **Recommended style**: [art] + [tone] (based on content signals)
-
-**Use AskUserQuestion** for:
-
-**Question 1: Visual Style**
-
-If a preset is recommended (see Auto Selection), show it first:
-
-```
-header: "Style"
-question: "Which visual style for this comic?"
-options:
-  - label: "[preset name] preset (Recommended)"       # If preset recommended
-    description: "[preset description] - includes special rules"
-  - label: "[recommended art] + [recommended tone] (Recommended)"  # If no preset
-    description: "Best match for your content based on analysis"
-  - label: "ligne-claire + neutral"
-    description: "Classic educational, Logicomix style"
-  - label: "ohmsha preset"
-    description: "Educational manga with visual metaphors, gadgets, NO talking heads"
-  - label: "Custom"
-    description: "Specify your own art + tone or preset"
-```
-
-**Preset vs Art+Tone**: Presets include special rules beyond art+tone. `ohmsha` = manga + neutral + visual metaphor rules + character roles + NO talking heads. Plain `manga + neutral` does NOT include these rules.
-
-**Question 2: Narrative Focus** (multiSelect: true)
-```
-header: "Focus"
-question: "What should the comic emphasize? (Select all that apply)"
-options:
-  - label: "Biography/life story"
-    description: "Follow a person's journey through key life events"
-  - label: "Concept explanation"
-    description: "Break down complex ideas visually"
-  - label: "Historical event"
-    description: "Dramatize important historical moments"
-  - label: "Tutorial/how-to"
-    description: "Step-by-step educational guide"
-```
-
-**Question 3: Target Audience**
-```
-header: "Audience"
-question: "Who is the primary reader?"
-options:
-  - label: "General readers"
-    description: "Broad appeal, accessible content"
-  - label: "Students/learners"
-    description: "Educational focus, clear explanations"
-  - label: "Industry professionals"
-    description: "Technical depth, domain knowledge"
-  - label: "Children/young readers"
-    description: "Simplified language, engaging visuals"
-```
-
-**Question 4: Outline Review**
-```
-header: "Review"
-question: "Do you want to review the outline before image generation?"
-options:
-  - label: "Yes, let me review (Recommended)"
-    description: "Review storyboard and characters before generating images"
-  - label: "No, generate directly"
-    description: "Skip outline review, start generating immediately"
-```
-
-**After response**:
-1. Update `analysis.md` with user preferences
-2. **Store `skip_outline_review`** flag based on Question 4 response
-3. → Step 3
-
-### Step 3: Generate Storyboard + Characters
-
-Create storyboard and character definitions using the confirmed style from Step 2.
-
-**Loading Style References**:
-- Art style: `references/art-styles/{art}.md`
-- Tone: `references/tones/{tone}.md`
-- If preset (ohmsha/wuxia/shoujo): also load `references/presets/{preset}.md`
-
-**Generate**:
-
-1. **Storyboard** (`storyboard.md`):
-   - YAML front matter with art_style, tone, layout, aspect_ratio
-   - Cover design
-   - Each page: layout, panel breakdown, visual prompts
-   - **Written in user's preferred language** (from Step 1)
-   - Reference: `references/storyboard-template.md`
-   - **If using preset**: Load and apply preset rules from `references/presets/`
-
-2. **Character definitions** (`characters/characters.md`):
-   - Visual specs matching the art style (in user's preferred language)
-   - Include Reference Sheet Prompt for later image generation
-   - Reference: `references/character-template.md`
-
-**After generation**:
-- If `skip_outline_review` is true → Skip Step 4, go directly to Step 5
-- If `skip_outline_review` is false → Continue to Step 4
-
-### Step 4: Confirmation 2 - Review Outline (Conditional)
-
-**Skip this step** if user selected "No, generate directly" in Step 2.
-
-**Purpose**: User reviews and confirms storyboard + characters before generation.
-
-**Display**:
-- Page count and structure
-- Art style + Tone combination
-- Page-by-page summary (Cover → P1 → P2...)
-- Character list with brief descriptions
-
-**Use AskUserQuestion**:
-
-```
-header: "Confirm"
-question: "Ready to generate images with this outline?"
-options:
-  - label: "Yes, proceed (Recommended)"
-    description: "Generate character sheet and comic pages"
-  - label: "Edit storyboard first"
-    description: "I'll modify storyboard.md before continuing"
-  - label: "Edit characters first"
-    description: "I'll modify characters/characters.md before continuing"
-  - label: "Edit both"
-    description: "I'll modify both files before continuing"
-```
-
-**After response**:
-1. If user wants to edit → Wait for user to finish editing, then ask again
-2. If user confirms → Continue to Step 5
-
-### Step 5: Generate Images
-
-With confirmed storyboard + art style + tone + aspect ratio:
-
-**Style Reference Loading**:
-- Read `references/art-styles/{art}.md` for rendering guidelines
-- Read `references/tones/{tone}.md` for mood/color adjustments
-- If preset: Read `references/presets/{preset}.md` for special rules
-
-**5.1 Generate Character Reference Sheet** (first):
-1. Use Reference Sheet Prompt from `characters/characters.md`
-2. Generate → `characters/characters.png`
-3. This ensures visual consistency for all subsequent pages
-
-**5.2 Generate Comic Pages**:
-
-**CRITICAL: Character Reference is MANDATORY** for visual consistency across all pages.
-
-**Before generating any page**:
-1. Read the image generation skill's SKILL.md
-2. Check if it supports reference image input (`--ref`, `--reference`, etc.)
-3. Choose the appropriate strategy below
-
-**Character Reference Strategy**:
-
-| Skill Capability | Strategy | Action |
-|------------------|----------|--------|
-| Supports `--ref` | **Strategy A** | Pass `characters/characters.png` with EVERY page |
-| Does NOT support `--ref` | **Strategy B** | Prepend character descriptions to EVERY prompt |
-
-**Strategy A: Using `--ref` parameter** (e.g., baoyu-image-gen)
-
-```bash
-# Each page generation MUST include --ref
+# Example: ALWAYS include --ref for consistency
 npx -y bun ${SKILL_DIR}/../baoyu-image-gen/scripts/main.ts \
   --promptfiles prompts/01-page-xxx.md \
-  --image 01-page-xxx.png \
-  --ar 3:4 \
+  --image 01-page-xxx.png --ar 3:4 \
   --ref characters/characters.png
 ```
 
-**Strategy B: Embedding character descriptions in prompt**
+**Full workflow details**: [references/workflow.md](references/workflow.md)
 
-When skill does NOT support reference images, create combined prompt files:
+### EXTEND.md Paths
 
-```markdown
-# prompts/01-page-xxx.md (with embedded character reference)
+| Path | Location |
+|------|----------|
+| `.baoyu-skills/baoyu-comic/EXTEND.md` | Project directory |
+| `$HOME/.baoyu-skills/baoyu-comic/EXTEND.md` | User home |
 
-## Character Reference (maintain consistency)
-[Copy relevant sections from characters/characters.md here]
-- 大雄: Japanese boy, round glasses, yellow shirt, navy shorts...
-- 哆啦A梦: Round blue robot cat, white belly, red nose, golden bell...
+**EXTEND.md Supports**: Watermark | Preferred art/tone/layout | Custom style definitions | Character presets | Language preference
 
-## Page Content
-[Original page prompt here]
-```
-
-**For each page (cover + pages)**:
-1. Save prompt to `prompts/NN-{cover|page}-[slug].md`
-2. Generate image using Strategy A or B (based on skill capability)
-3. Report progress after each generation
-
-**Watermark Application** (if enabled in preferences):
-Add to each image generation prompt:
-```
-Include a subtle watermark "[content]" positioned at [position]
-with approximately [opacity*100]% visibility. The watermark should
-be legible but not distracting from the comic panels and storytelling.
-Ensure watermark does not overlap speech bubbles or key action.
-```
-Reference: `references/config/watermark-guide.md`
-
-**Session Management**:
-If image generation skill supports `--sessionId`:
-1. Generate unique session ID: `comic-{topic-slug}-{timestamp}`
-2. Use same session ID for all pages
-3. Ensures visual consistency across generated images
-
-### Step 6: Merge to PDF
-
-After all images generated:
-
-```bash
-npx -y bun ${SKILL_DIR}/scripts/merge-to-pdf.ts <comic-dir>
-```
-
-Creates `{topic-slug}.pdf` with all pages as full-page images.
-
-### Step 7: Completion Report
-
-```
-Comic Complete!
-Title: [title] | Art: [art] | Tone: [tone] | Pages: [count] | Aspect: [ratio] | Language: [lang]
-Watermark: [enabled/disabled]
-Location: [path]
-✓ analysis.md
-✓ characters.png
-✓ 00-cover-[slug].png ... NN-page-[slug].png
-✓ {topic-slug}.pdf
-```
-
-## Page Modification
-
-| Action | Steps |
-|--------|-------|
-| **Edit** | Update prompt → Regenerate image → Regenerate PDF |
-| **Add** | Create prompt at position → Generate image → Renumber subsequent (NN+1) → Update storyboard → Regenerate PDF |
-| **Delete** | Remove files → Renumber subsequent (NN-1) → Update storyboard → Regenerate PDF |
-
-**File naming**: `NN-{cover|page}-[slug].png` (e.g., `03-page-enigma-machine.png`)
-- Slugs: kebab-case, unique, derived from content
-- Renumbering: Update NN prefix only, slugs unchanged
-
-## Preset: Ohmsha
-
-Default: Use Doraemon characters (大雄=learner, 哆啦A梦=mentor, 胖虎=challenge, 静香=support). Custom characters via `--characters "Student:小明,Mentor:教授"` or character presets in EXTEND.md.
-
-Requirements: Visual metaphors, NO talking heads, narrative page titles.
-
-See `references/presets/ohmsha.md` and `references/ohmsha-guide.md` for details.
+Schema: [references/config/preferences-schema.md](references/config/preferences-schema.md)
 
 ## References
 
-Detailed templates and guidelines in `references/` directory:
-
 **Core Templates**:
-- `analysis-framework.md` - Deep content analysis for comic adaptation
-- `character-template.md` - Character definition format and examples
-- `storyboard-template.md` - Storyboard structure and panel breakdown
-- `ohmsha-guide.md` - Ohmsha manga style specifics
+- [analysis-framework.md](references/analysis-framework.md) - Deep content analysis
+- [character-template.md](references/character-template.md) - Character definition format
+- [storyboard-template.md](references/storyboard-template.md) - Storyboard structure
+- [ohmsha-guide.md](references/ohmsha-guide.md) - Ohmsha manga specifics
 
-**Style Definitions** (new 3-dimension system):
-- `art-styles/` - Art style definitions (ligne-claire, manga, realistic, ink-brush, chalk)
-- `tones/` - Tone definitions (neutral, warm, dramatic, romantic, energetic, vintage, action)
-- `presets/` - Preset shortcuts with special rules (ohmsha, wuxia, shoujo)
-- `layouts/` - Layout definitions (standard, cinematic, dense, splash, mixed, webtoon)
+**Style Definitions**:
+- `references/art-styles/` - Art styles (ligne-claire, manga, realistic, ink-brush, chalk)
+- `references/tones/` - Tones (neutral, warm, dramatic, romantic, energetic, vintage, action)
+- `references/presets/` - Presets with special rules (ohmsha, wuxia, shoujo)
+- `references/layouts/` - Layouts (standard, cinematic, dense, splash, mixed, webtoon)
+
+**Workflow**:
+- [workflow.md](references/workflow.md) - Full workflow details
+- [auto-selection.md](references/auto-selection.md) - Content signal analysis
+- [partial-workflows.md](references/partial-workflows.md) - Partial workflow options
 
 **Config**:
-- `config/preferences-schema.md` - EXTEND.md schema
-- `config/first-time-setup.md` - First-time setup flow
-- `config/watermark-guide.md` - Watermark configuration
+- [config/preferences-schema.md](references/config/preferences-schema.md) - EXTEND.md schema
+- [config/first-time-setup.md](references/config/first-time-setup.md) - First-time setup
+- [config/watermark-guide.md](references/config/watermark-guide.md) - Watermark configuration
 
 ## Notes
 
-- Auto-retry once on failure | Cartoon alternatives for sensitive figures
-- Use confirmed language preference | Maintain style consistency
-- **Step 2 confirmation required** - do not skip (style selection + outline review option)
-- **Step 4 conditional** - only if user requested outline review in Step 2
-- Watermark/language configured once in EXTEND.md, not asked in confirmations
-
-## Extension Support
-
-Custom configurations via EXTEND.md. See **Step 1.1** for paths and supported options.
+- Image generation: 10-30 seconds per page
+- Auto-retry once on generation failure
+- Use stylized alternatives for sensitive public figures
+- Maintain style consistency via session ID
+- **Step 2 confirmation required** - do not skip
+- **Steps 4/6 conditional** - only if user requested in Step 2
+- **Step 7.1 character sheet MUST be generated before pages** - ensures consistency
+- **Step 7.2 EVERY page MUST reference characters** - use `--ref` or embed descriptions
+- Watermark/language configured once in EXTEND.md
diff --git a/skills/baoyu-comic/references/auto-selection.md b/skills/baoyu-comic/references/auto-selection.md
new file mode 100644
index 0000000..3ad593a
--- /dev/null
+++ b/skills/baoyu-comic/references/auto-selection.md
@@ -0,0 +1,58 @@
+# Auto Selection
+
+Content signals determine default art + tone + layout (or preset).
+
+## Content Signal Matrix
+
+| Content Signals | Art Style | Tone | Layout | Preset |
+|-----------------|-----------|------|--------|--------|
+| Tutorial, how-to, beginner | manga | neutral | webtoon | **ohmsha** |
+| Computing, AI, programming | manga | neutral | dense | **ohmsha** |
+| Technical explanation, educational | manga | neutral | webtoon | **ohmsha** |
+| Pre-1950, classical, ancient | realistic | vintage | cinematic | - |
+| Personal story, mentor | ligne-claire | warm | standard | - |
+| Conflict, breakthrough | (inherit) | dramatic | splash | - |
+| Wine, food, business, lifestyle | realistic | neutral | cinematic | - |
+| Martial arts, wuxia, xianxia | ink-brush | action | splash | **wuxia** |
+| Romance, love, school life | manga | romantic | standard | **shoujo** |
+| Biography, balanced | ligne-claire | neutral | mixed | - |
+
+## Preset Recommendation Rules
+
+**When preset is recommended**: Load `presets/{preset}.md` and apply all special rules.
+
+### ohmsha
+- **Triggers**: Tutorial, technical, educational, computing, programming, how-to, beginner
+- **Special rules**: Visual metaphors, NO talking heads, gadget reveals, Doraemon-style characters
+- **Base**: manga + neutral + webtoon/dense
+
+### wuxia
+- **Triggers**: Martial arts, wuxia, xianxia, cultivation, swordplay
+- **Special rules**: Qi effects, combat visuals, atmospheric elements
+- **Base**: ink-brush + action + splash
+
+### shoujo
+- **Triggers**: Romance, love story, school life, emotional drama
+- **Special rules**: Decorative elements, eye details, romantic beats
+- **Base**: manga + romantic + standard
+
+## Compatibility Matrix
+
+Art Style × Tone combinations work best when matched appropriately:
+
+| Art Style | ✓✓ Best | ✓ Works | ✗ Avoid |
+|-----------|---------|---------|---------|
+| ligne-claire | neutral, warm | dramatic, vintage, energetic | romantic, action |
+| manga | neutral, romantic, energetic, action | warm, dramatic | vintage |
+| realistic | neutral, warm, dramatic, vintage | action | romantic, energetic |
+| ink-brush | neutral, dramatic, action, vintage | warm | romantic, energetic |
+| chalk | neutral, warm, energetic | vintage | dramatic, action, romantic |
+
+**Note**: Art Style × Tone × Layout can be freely combined. Incompatible combinations work but may produce unexpected results.
+
+## Priority Order
+
+1. User-specified options (`--art`, `--tone`, `--style`)
+2. EXTEND.md defaults
+3. Content signal analysis → auto-selection
+4. Fallback: ligne-claire + neutral + standard
diff --git a/skills/baoyu-comic/references/partial-workflows.md b/skills/baoyu-comic/references/partial-workflows.md
new file mode 100644
index 0000000..ae04943
--- /dev/null
+++ b/skills/baoyu-comic/references/partial-workflows.md
@@ -0,0 +1,123 @@
+# Partial Workflows
+
+Options to run specific parts of the workflow.
+
+## Options Summary
+
+| Option | Steps Executed | Output |
+|--------|----------------|--------|
+| `--storyboard-only` | 1-3 | `storyboard.md` + `characters/` |
+| `--prompts-only` | 1-5 | + `prompts/*.md` |
+| `--images-only` | 7-9 | + images + PDF |
+| `--regenerate N` | 7 (partial) | Specific page(s) + PDF |
+
+---
+
+## Using `--storyboard-only`
+
+Generate storyboard and characters without prompts or images:
+
+```bash
+/baoyu-comic content.md --storyboard-only
+```
+
+**Workflow**: Steps 1-3 only (stop after storyboard + characters)
+
+**Output**:
+- `analysis.md`
+- `storyboard.md`
+- `characters/characters.md`
+
+**Use case**: Review and edit the storyboard before generating images. Useful for:
+- Getting feedback on the narrative structure
+- Making manual adjustments to panel layouts
+- Defining custom characters
+
+---
+
+## Using `--prompts-only`
+
+Generate storyboard, characters, and prompts without images:
+
+```bash
+/baoyu-comic content.md --prompts-only
+```
+
+**Workflow**: Steps 1-5 (generate prompts, skip images)
+
+**Output**:
+- `analysis.md`
+- `storyboard.md`
+- `characters/characters.md`
+- `prompts/*.md`
+
+**Use case**: Review and edit prompts before image generation. Useful for:
+- Fine-tuning image generation prompts
+- Ensuring visual consistency before committing to generation
+- Making style adjustments at the prompt level
+
+---
+
+## Using `--images-only`
+
+Generate images from existing prompts (starts at Step 7):
+
+```bash
+/baoyu-comic comic/topic-slug/ --images-only
+```
+
+**Workflow**: Skip to Step 7, then 8-9
+
+**Prerequisites** (must exist in directory):
+- `prompts/` directory with page prompt files
+- `storyboard.md` with style information
+- `characters/characters.md` with character definitions
+
+**Output**:
+- `characters/characters.png` (if not exists)
+- `NN-{cover|page}-[slug].png` images
+- `{topic-slug}.pdf`
+
+**Use case**: Re-generate images after editing prompts. Useful for:
+- Recovering from failed image generation
+- Trying different image generation settings
+- Regenerating after manual prompt edits
+
+---
+
+## Using `--regenerate`
+
+Regenerate specific pages only:
+
+```bash
+# Single page
+/baoyu-comic comic/topic-slug/ --regenerate 3
+
+# Multiple pages
+/baoyu-comic comic/topic-slug/ --regenerate 2,5,8
+
+# Cover page
+/baoyu-comic comic/topic-slug/ --regenerate 0
+```
+
+**Workflow**:
+1. Read existing prompts for specified pages
+2. Regenerate images only for those pages
+3. Regenerate PDF
+
+**Prerequisites** (must exist):
+- `prompts/NN-{cover|page}-[slug].md` for specified pages
+- `characters/characters.png` (for reference)
+
+**Output**:
+- Regenerated `NN-{cover|page}-[slug].png` for specified pages
+- Updated `{topic-slug}.pdf`
+
+**Use case**: Fix specific pages without regenerating entire comic. Useful for:
+- Fixing a single problematic page
+- Iterating on specific visuals
+- Regenerating pages after prompt edits
+
+**Page numbering**:
+- `0` = Cover page
+- `1-N` = Content pages
diff --git a/skills/baoyu-comic/references/presets/ohmsha.md b/skills/baoyu-comic/references/presets/ohmsha.md
index 8609cf9..2ff392c 100644
--- a/skills/baoyu-comic/references/presets/ohmsha.md
+++ b/skills/baoyu-comic/references/presets/ohmsha.md
@@ -41,16 +41,18 @@ Every technical concept MUST be visualized as a metaphor:
 
 ### Character Roles (Required)
 
-Define characters before generating:
+**DEFAULT: Use Doraemon characters** unless user explicitly specifies `--characters` or has character presets in EXTEND.md.
 
-| Role | Default | Traits |
-|------|---------|--------|
-| Student (Role A) | 大雄 | Confused, asks basic but crucial questions, represents reader |
-| Mentor (Role B) | 哆啦A梦 | Knowledgeable, patient, uses gadgets as technical metaphors |
-| Challenge (Role C, optional) | 胖虎 | Represents misunderstanding, or "noise" in the data |
-| Support (Role D, optional) | 静香 | Asks clarifying questions, provides alternative perspectives |
+| Role | Default Character | Visual | Traits |
+|------|-------------------|--------|--------|
+| Student (Role A) | 大雄 (Nobita) | Boy, 10yo, round glasses, black hair, yellow shirt, navy shorts | Confused, asks basic but crucial questions, represents reader |
+| Mentor (Role B) | 哆啦A梦 (Doraemon) | Blue robot cat, white belly, 4D pocket, red nose, golden bell | Knowledgeable, patient, uses gadgets as technical metaphors |
+| Challenge (Role C) | 胖虎 (Gian) | Stocky boy, small eyes, orange shirt | Represents misunderstanding, or "noise" in the data |
+| Support (Role D) | 静香 (Shizuka) | Cute girl, black short hair, pink dress | Asks clarifying questions, provides alternative perspectives |
 
-Custom characters via `--characters "Student:小明,Mentor:教授"` or EXTEND.md presets.
+**IMPORTANT**: These Doraemon characters ARE the default for ohmsha preset. Generate character definitions using these exact characters unless user requests otherwise.
+
+To use custom characters: `--characters "Student:小明,Mentor:教授"` or define in EXTEND.md.
 
 ### Page Title Convention
 
diff --git a/skills/baoyu-comic/references/workflow.md b/skills/baoyu-comic/references/workflow.md
new file mode 100644
index 0000000..ce2df39
--- /dev/null
+++ b/skills/baoyu-comic/references/workflow.md
@@ -0,0 +1,505 @@
+# Complete Workflow
+
+Full workflow for generating knowledge comics.
+
+## Progress Checklist
+
+Copy and track progress:
+
+```
+Comic Progress:
+- [ ] Step 1: Setup & Analyze
+  - [ ] 1.1 Load preferences
+  - [ ] 1.2 Analyze content
+  - [ ] 1.3 Check existing ⚠️ REQUIRED
+- [ ] Step 2: Confirmation 1 - Style & options ⚠️ REQUIRED
+- [ ] Step 3: Generate storyboard + characters
+- [ ] Step 4: Review outline (conditional)
+- [ ] Step 5: Generate prompts
+- [ ] Step 6: Review prompts (conditional)
+- [ ] Step 7: Generate images
+- [ ] Step 8: Merge to PDF
+- [ ] Step 9: Completion report
+```
+
+## Flow Diagram
+
+```
+Input → Preferences → Analyze → [Check Existing?] → [Confirm 1: Style + Reviews] → Storyboard → [Review Outline?] → Prompts → [Review Prompts?] → Images → PDF → Complete
+```
+
+---
+
+## Step 1: Setup & Analyze
+
+### 1.1 Load Preferences (EXTEND.md)
+
+Use Bash to check EXTEND.md existence (priority order):
+
+```bash
+# Check project-level first
+test -f .baoyu-skills/baoyu-comic/EXTEND.md && echo "project"
+
+# Then user-level (cross-platform: $HOME works on macOS/Linux/WSL)
+test -f "$HOME/.baoyu-skills/baoyu-comic/EXTEND.md" && echo "user"
+```
+
+| Path | Location |
+|------|----------|
+| `.baoyu-skills/baoyu-comic/EXTEND.md` | Project directory |
+| `$HOME/.baoyu-skills/baoyu-comic/EXTEND.md` | User home |
+
+**When EXTEND.md Found** → Read, parse, **output summary to user**:
+
+```
+📋 Loaded preferences from [full path]
+├─ Watermark: [enabled/disabled] [content if enabled]
+├─ Art Style: [style name or "auto-select"]
+├─ Tone: [tone name or "auto-select"]
+├─ Layout: [layout or "auto-select"]
+├─ Language: [language or "auto-detect"]
+└─ Character presets: [count] defined
+```
+
+**MUST output this summary** so user knows their current configuration. Do not skip or silently load.
+
+**When EXTEND.md Not Found** → First-time setup:
+
+1. Inform user: "No preferences found. Let's set up your defaults."
+2. Use AskUserQuestion to collect preferences (see `config/first-time-setup.md`)
+3. Create EXTEND.md at user-chosen location
+4. Confirm: "✓ Preferences saved to [path]"
+
+**EXTEND.md Supports**: Watermark | Preferred art/tone/layout | Custom style definitions | Character presets | Language preference
+
+Schema: `config/preferences-schema.md`
+
+**Important**: Once EXTEND.md exists, watermark, language, and style defaults are NOT asked again in Confirmation 1 or 2. These are session-persistent settings.
+
+### 1.2 Analyze Content → `analysis.md`
+
+Read source content, save it if needed, and perform deep analysis.
+
+**Actions**:
+1. **Save source content** (if not already a file):
+   - If user provides a file path: use as-is
+   - If user pastes content: save to `source.md` in target directory
+2. Read source content
+3. **Deep analysis** following `analysis-framework.md`:
+   - Target audience identification
+   - Value proposition for readers
+   - Core themes and narrative potential
+   - Key figures and their story arcs
+4. Detect source language
+5. **Determine language**:
+   - If EXTEND.md has `language` → use it
+   - Else if `--lang` option provided → use it
+   - Else → use detected source language
+6. Determine recommended page count:
+   - Short story: 5-8 pages
+   - Medium complexity: 9-15 pages
+   - Full biography: 16-25 pages
+7. Analyze content signals for art/tone/layout recommendations
+8. **Save to `analysis.md`**
+
+**analysis.md Format**: YAML front matter (title, topic, time_span, source_language, user_language, aspect_ratio, recommended_page_count, recommended_art, recommended_tone) + sections for Target Audience, Value Proposition, Core Themes, Key Figures & Story Arcs, Content Signals, Recommended Approaches. See `analysis-framework.md` for full template.
+
+### 1.3 Check Existing Content ⚠️ REQUIRED
+
+**MUST execute before proceeding to Step 2.**
+
+Use Bash to check if output directory exists:
+
+```bash
+test -d "comic/{topic-slug}" && echo "exists"
+```
+
+**If directory exists**, use AskUserQuestion:
+
+```
+header: "Existing"
+question: "Existing content found. How to proceed?"
+options:
+  - label: "Regenerate storyboard"
+    description: "Keep images, regenerate storyboard and characters only"
+  - label: "Regenerate images"
+    description: "Keep storyboard, regenerate images only"
+  - label: "Backup and regenerate"
+    description: "Backup to {slug}-backup-{timestamp}, then regenerate all"
+  - label: "Exit"
+    description: "Cancel, keep existing content unchanged"
+```
+
+Save result and handle accordingly:
+- **Regenerate storyboard**: Skip to Step 3, preserve `prompts/` and images
+- **Regenerate images**: Skip to Step 7, use existing prompts
+- **Backup and regenerate**: Move directory, start fresh from Step 2
+- **Exit**: End workflow immediately
+
+---
+
+## Step 2: Confirmation 1 - Style & Options ⚠️
+
+**Purpose**: Select visual style + decide whether to review outline before generation. **Do NOT skip.**
+
+**Note**: Watermark and language already configured in EXTEND.md (Step 1).
+
+**Display summary**:
+- Content type + topic identified
+- Key figures extracted
+- Time span detected
+- Recommended page count
+- Language: [from EXTEND.md or detected]
+- **Recommended style**: [art] + [tone] (based on content signals)
+
+**Use AskUserQuestion** for:
+
+### Question 1: Visual Style
+
+If a preset is recommended (see `auto-selection.md`), show it first:
+
+```
+header: "Style"
+question: "Which visual style for this comic?"
+options:
+  - label: "[preset name] preset (Recommended)"       # If preset recommended
+    description: "[preset description] - includes special rules"
+  - label: "[recommended art] + [recommended tone] (Recommended)"  # If no preset
+    description: "Best match for your content based on analysis"
+  - label: "ligne-claire + neutral"
+    description: "Classic educational, Logicomix style"
+  - label: "ohmsha preset"
+    description: "Educational manga with visual metaphors, gadgets, NO talking heads"
+  - label: "Custom"
+    description: "Specify your own art + tone or preset"
+```
+
+**Preset vs Art+Tone**: Presets include special rules beyond art+tone. `ohmsha` = manga + neutral + visual metaphor rules + character roles + NO talking heads. Plain `manga + neutral` does NOT include these rules.
+
+### Question 2: Narrative Focus (multiSelect: true)
+
+```
+header: "Focus"
+question: "What should the comic emphasize? (Select all that apply)"
+options:
+  - label: "Biography/life story"
+    description: "Follow a person's journey through key life events"
+  - label: "Concept explanation"
+    description: "Break down complex ideas visually"
+  - label: "Historical event"
+    description: "Dramatize important historical moments"
+  - label: "Tutorial/how-to"
+    description: "Step-by-step educational guide"
+```
+
+### Question 3: Target Audience
+
+```
+header: "Audience"
+question: "Who is the primary reader?"
+options:
+  - label: "General readers"
+    description: "Broad appeal, accessible content"
+  - label: "Students/learners"
+    description: "Educational focus, clear explanations"
+  - label: "Industry professionals"
+    description: "Technical depth, domain knowledge"
+  - label: "Children/young readers"
+    description: "Simplified language, engaging visuals"
+```
+
+### Question 4: Outline Review
+
+```
+header: "Review"
+question: "Do you want to review the outline before image generation?"
+options:
+  - label: "Yes, let me review (Recommended)"
+    description: "Review storyboard and characters before generating images"
+  - label: "No, generate directly"
+    description: "Skip outline review, start generating immediately"
+```
+
+### Question 5: Prompt Review
+
+```
+header: "Prompts"
+question: "Review prompts before generating images?"
+options:
+  - label: "Yes, review prompts (Recommended)"
+    description: "Review image generation prompts before generating"
+  - label: "No, skip prompt review"
+    description: "Proceed directly to image generation"
+```
+
+**After response**:
+1. Update `analysis.md` with user preferences
+2. **Store `skip_outline_review`** flag based on Question 4 response
+3. **Store `skip_prompt_review`** flag based on Question 5 response
+4. → Step 3
+
+---
+
+## Step 3: Generate Storyboard + Characters
+
+Create storyboard and character definitions using the confirmed style from Step 2.
+
+**Loading Style References**:
+- Art style: `art-styles/{art}.md`
+- Tone: `tones/{tone}.md`
+- If preset (ohmsha/wuxia/shoujo): also load `presets/{preset}.md`
+
+**Generate**:
+
+1. **Storyboard** (`storyboard.md`):
+   - YAML front matter with art_style, tone, layout, aspect_ratio
+   - Cover design
+   - Each page: layout, panel breakdown, visual prompts
+   - **Written in user's preferred language** (from Step 1)
+   - Reference: `storyboard-template.md`
+   - **If using preset**: Load and apply preset rules from `presets/`
+
+2. **Character definitions** (`characters/characters.md`):
+   - Visual specs matching the art style (in user's preferred language)
+   - Include Reference Sheet Prompt for later image generation
+   - Reference: `character-template.md`
+   - **If using ohmsha preset**: Use default Doraemon characters (see below)
+
+**Ohmsha Default Characters** (use these unless user specifies `--characters`):
+
+| Role | Character | Visual Description |
+|------|-----------|-------------------|
+| Student | 大雄 (Nobita) | Japanese boy, 10yo, round glasses, black hair parted in middle, yellow shirt, navy shorts |
+| Mentor | 哆啦A梦 (Doraemon) | Round blue robot cat, big white eyes, red nose, whiskers, white belly with 4D pocket, golden bell, no ears |
+| Challenge | 胖虎 (Gian) | Stocky boy, rough features, small eyes, orange shirt |
+| Support | 静香 (Shizuka) | Cute girl, black short hair, pink dress, gentle expression |
+
+These are the canonical ohmsha-style characters. Do NOT create custom characters for ohmsha unless explicitly requested.
+
+**After generation**:
+- If `skip_outline_review` is true → Skip Step 4, go directly to Step 5
+- If `skip_outline_review` is false → Continue to Step 4
+
+---
+
+## Step 4: Review Outline (Conditional)
+
+**Skip this step** if user selected "No, generate directly" in Step 2.
+
+**Purpose**: User reviews and confirms storyboard + characters before generation.
+
+**Display**:
+- Page count and structure
+- Art style + Tone combination
+- Page-by-page summary (Cover → P1 → P2...)
+- Character list with brief descriptions
+
+**Use AskUserQuestion**:
+
+```
+header: "Confirm"
+question: "Ready to generate images with this outline?"
+options:
+  - label: "Yes, proceed (Recommended)"
+    description: "Generate character sheet and comic pages"
+  - label: "Edit storyboard first"
+    description: "I'll modify storyboard.md before continuing"
+  - label: "Edit characters first"
+    description: "I'll modify characters/characters.md before continuing"
+  - label: "Edit both"
+    description: "I'll modify both files before continuing"
+```
+
+**After response**:
+1. If user wants to edit → Wait for user to finish editing, then ask again
+2. If user confirms → Continue to Step 5
+
+---
+
+## Step 5: Generate Prompts
+
+Create image generation prompts for all pages.
+
+**Style Reference Loading**:
+- Read `art-styles/{art}.md` for rendering guidelines
+- Read `tones/{tone}.md` for mood/color adjustments
+- If preset: Read `presets/{preset}.md` for special rules
+
+**For each page (cover + pages)**:
+1. Create prompt following art style + tone guidelines
+2. Include character visual descriptions for consistency
+3. Save to `prompts/NN-{cover|page}-[slug].md`
+
+**Prompt File Format**:
+```markdown
+# Page NN: [Title]
+
+## Visual Style
+Art: [art style] | Tone: [tone] | Layout: [layout type]
+
+## Character Reference
+[Character descriptions from characters/characters.md]
+
+## Panel Breakdown
+[From storyboard.md - panel descriptions, actions, dialogue]
+
+## Generation Prompt
+[Combined prompt for image generation skill]
+```
+
+**Watermark Application** (if enabled in preferences):
+Add to each prompt:
+```
+Include a subtle watermark "[content]" positioned at [position]
+with approximately [opacity*100]% visibility. The watermark should
+be legible but not distracting from the comic panels and storytelling.
+Ensure watermark does not overlap speech bubbles or key action.
+```
+Reference: `config/watermark-guide.md`
+
+**After generation**:
+- If `skip_prompt_review` is true → Skip Step 6, go directly to Step 7
+- If `skip_prompt_review` is false → Continue to Step 6
+
+---
+
+## Step 6: Review Prompts (Conditional)
+
+**Skip this step** if user selected "No, skip prompt review" in Step 2.
+
+**Purpose**: User reviews and confirms prompts before image generation.
+
+**Display prompt summary table**:
+
+| Page | Title | Key Elements |
+|------|-------|--------------|
+| Cover | [title] | [main visual] |
+| P1 | [title] | [key elements] |
+| ... | ... | ... |
+
+**Use AskUserQuestion**:
+
+```
+header: "Confirm"
+question: "Ready to generate images with these prompts?"
+options:
+  - label: "Yes, proceed (Recommended)"
+    description: "Generate all comic page images"
+  - label: "Edit prompts first"
+    description: "I'll modify prompts/*.md before continuing"
+  - label: "Regenerate prompts"
+    description: "Regenerate all prompts with different approach"
+```
+
+**After response**:
+1. If user wants to edit → Wait for user to finish editing, then ask again
+2. If user wants to regenerate → Go back to Step 5
+3. If user confirms → Continue to Step 7
+
+---
+
+## Step 7: Generate Images
+
+With confirmed prompts from Step 5/6:
+
+### 7.1 Generate Character Reference Sheet (first)
+
+1. Use Reference Sheet Prompt from `characters/characters.md`
+2. Generate → `characters/characters.png`
+3. This ensures visual consistency for all subsequent pages
+
+### 7.2 Generate Comic Pages
+
+**CRITICAL: Character Reference is MANDATORY** for visual consistency across all pages.
+
+**Before generating any page**:
+1. Read the image generation skill's SKILL.md
+2. Check if it supports reference image input (`--ref`, `--reference`, etc.)
+3. Choose the appropriate strategy below
+
+**Character Reference Strategy**:
+
+| Skill Capability | Strategy | Action |
+|------------------|----------|--------|
+| Supports `--ref` | **Strategy A** | Pass `characters/characters.png` with EVERY page |
+| Does NOT support `--ref` | **Strategy B** | Prepend character descriptions to EVERY prompt |
+
+**Strategy A: Using `--ref` parameter** (e.g., baoyu-image-gen)
+
+```bash
+# Each page generation MUST include --ref
+npx -y bun ${SKILL_DIR}/../baoyu-image-gen/scripts/main.ts \
+  --promptfiles prompts/01-page-xxx.md \
+  --image 01-page-xxx.png \
+  --ar 3:4 \
+  --ref characters/characters.png
+```
+
+**Strategy B: Embedding character descriptions in prompt**
+
+When skill does NOT support reference images, create combined prompt files:
+
+```markdown
+# prompts/01-page-xxx.md (with embedded character reference)
+
+## Character Reference (maintain consistency)
+[Copy relevant sections from characters/characters.md here]
+- 大雄: Japanese boy, round glasses, yellow shirt, navy shorts...
+- 哆啦A梦: Round blue robot cat, white belly, red nose, golden bell...
+
+## Page Content
+[Original page prompt here]
+```
+
+**For each page (cover + pages)**:
+1. Read prompt from `prompts/NN-{cover|page}-[slug].md`
+2. Generate image using Strategy A or B (based on skill capability)
+3. Save to `NN-{cover|page}-[slug].png`
+4. Report progress after each generation: "Generated X/N: [page title]"
+
+**Session Management**:
+If image generation skill supports `--sessionId`:
+1. Generate unique session ID: `comic-{topic-slug}-{timestamp}`
+2. Use same session ID for all pages
+3. Ensures visual consistency across generated images
+
+---
+
+## Step 8: Merge to PDF
+
+After all images generated:
+
+```bash
+npx -y bun ${SKILL_DIR}/scripts/merge-to-pdf.ts <comic-dir>
+```
+
+Creates `{topic-slug}.pdf` with all pages as full-page images.
+
+---
+
+## Step 9: Completion Report
+
+```
+Comic Complete!
+Title: [title] | Art: [art] | Tone: [tone] | Pages: [count] | Aspect: [ratio] | Language: [lang]
+Watermark: [enabled/disabled]
+Location: [path]
+✓ analysis.md
+✓ characters.png
+✓ 00-cover-[slug].png ... NN-page-[slug].png
+✓ {topic-slug}.pdf
+```
+
+---
+
+## Page Modification
+
+| Action | Steps |
+|--------|-------|
+| **Edit** | Update prompt → Regenerate image → Regenerate PDF |
+| **Add** | Create prompt at position → Generate image → Renumber subsequent (NN+1) → Update storyboard → Regenerate PDF |
+| **Delete** | Remove files → Renumber subsequent (NN-1) → Update storyboard → Regenerate PDF |
+
+**File naming**: `NN-{cover|page}-[slug].png` (e.g., `03-page-enigma-machine.png`)
+- Slugs: kebab-case, unique, derived from content
+- Renumbering: Update NN prefix only, slugs unchanged
diff --git a/skills/baoyu-image-gen/SKILL.md b/skills/baoyu-image-gen/SKILL.md
index 758f0dd..8668460 100644
--- a/skills/baoyu-image-gen/SKILL.md
+++ b/skills/baoyu-image-gen/SKILL.md
@@ -76,7 +76,8 @@ npx -y bun ${SKILL_DIR}/scripts/main.ts --prompt "A cat" --image out.png --provi
 | `--model <id>`, `-m` | Model ID |
 | `--ar <ratio>` | Aspect ratio (e.g., `16:9`, `1:1`, `4:3`) |
 | `--size <WxH>` | Size (e.g., `1024x1024`) |
-| `--quality normal\|2k` | Quality preset (default: normal) |
+| `--quality normal\|2k` | Quality preset (default: 2k) |
+| `--imageSize 1K\|2K\|4K` | Image size for Google (default: from quality) |
 | `--ref <files...>` | Reference images (Google multimodal only) |
 | `--n <count>` | Number of images |
 | `--json` | JSON output |
@@ -102,17 +103,20 @@ npx -y bun ${SKILL_DIR}/scripts/main.ts --prompt "A cat" --image out.png --provi
 
 ## Quality Presets
 
-| Preset | Resolution | Use Case |
-|--------|------------|----------|
-| `normal` | ~1024px | Covers, illustrations |
-| `2k` | ~2048px | Infographics, slides |
+| Preset | Google imageSize | OpenAI Size | Use Case |
+|--------|------------------|-------------|----------|
+| `normal` | 1K | 1024px | Quick previews |
+| `2k` (default) | 2K | 2048px | Covers, illustrations, infographics |
+
+**Google imageSize**: Can be overridden with `--imageSize 1K|2K|4K`
 
 ## Aspect Ratios
 
 Supported: `1:1`, `16:9`, `9:16`, `4:3`, `3:4`, `2.35:1`
 
-- Multimodal models: embedded in prompt
-- Image-only models: uses `aspectRatio` or `size` parameter
+- Google multimodal: uses `imageConfig.aspectRatio`
+- Google Imagen: uses `aspectRatio` parameter
+- OpenAI: maps to closest supported size
 
 ## Error Handling
 
diff --git a/skills/baoyu-image-gen/scripts/main.ts b/skills/baoyu-image-gen/scripts/main.ts
index ed46b0d..88bd331 100644
--- a/skills/baoyu-image-gen/scripts/main.ts
+++ b/skills/baoyu-image-gen/scripts/main.ts
@@ -1,34 +1,8 @@
-import fs from "node:fs";
 import path from "node:path";
 import process from "node:process";
 import { homedir } from "node:os";
 import { mkdir, readFile, writeFile } from "node:fs/promises";
-
-type Provider = "google" | "openai";
-type Quality = "normal" | "2k";
-
-type CliArgs = {
-  prompt: string | null;
-  promptFiles: string[];
-  imagePath: string | null;
-  provider: Provider | null;
-  model: string | null;
-  aspectRatio: string | null;
-  size: string | null;
-  quality: Quality;
-  referenceImages: string[];
-  n: number;
-  json: boolean;
-  help: boolean;
-};
-
-const GOOGLE_MULTIMODAL_MODELS = [
-  "gemini-3-pro-image-preview",
-];
-
-const GOOGLE_IMAGEN_MODELS = ["imagen-3.0-generate-002", "imagen-3.0-generate-001"];
-
-const OPENAI_IMAGE_MODELS = ["gpt-image-1.5", "gpt-image-1", "dall-e-3", "dall-e-2"];
+import type { CliArgs, Provider } from "./types";
 
 function printUsage(): void {
   console.log(`Usage:
@@ -44,7 +18,8 @@ Options:
   -m, --model <id>          Model ID
   --ar <ratio>              Aspect ratio (e.g., 16:9, 1:1, 4:3)
   --size <WxH>              Size (e.g., 1024x1024)
-  --quality normal|2k       Quality preset (default: normal)
+  --quality normal|2k       Quality preset (default: 2k)
+  --imageSize 1K|2K|4K      Image size for Google (default: from quality)
   --ref <files...>          Reference images (Google multimodal only)
   --n <count>               Number of images (default: 1)
   --json                    JSON output
@@ -53,6 +28,7 @@ Options:
 Environment variables:
   OPENAI_API_KEY            OpenAI API key
   GOOGLE_API_KEY            Google API key
+  GEMINI_API_KEY            Gemini API key (alias for GOOGLE_API_KEY)
   OPENAI_IMAGE_MODEL        Default OpenAI model (gpt-image-1.5)
   GOOGLE_IMAGE_MODEL        Default Google model (gemini-3-pro-image-preview)
   OPENAI_BASE_URL           Custom OpenAI endpoint
@@ -70,7 +46,8 @@ function parseArgs(argv: string[]): CliArgs {
     model: null,
     aspectRatio: null,
     size: null,
-    quality: "normal",
+    quality: "2k",
+    imageSize: null,
     referenceImages: [],
     n: 1,
     json: false,
@@ -161,6 +138,13 @@ function parseArgs(argv: string[]): CliArgs {
       continue;
     }
 
+    if (a === "--imageSize") {
+      const v = argv[++i]?.toUpperCase();
+      if (v !== "1K" && v !== "2K" && v !== "4K") throw new Error(`Invalid imageSize: ${v}`);
+      out.imageSize = v;
+      continue;
+    }
+
     if (a === "--ref" || a === "--reference") {
       const { items, next } = takeMany(i);
       if (items.length === 0) throw new Error(`Missing files for ${a}`);
@@ -257,7 +241,7 @@ function normalizeOutputImagePath(p: string): string {
 function detectProvider(args: CliArgs): Provider {
   if (args.provider) return args.provider;
 
-  const hasGoogle = !!process.env.GOOGLE_API_KEY;
+  const hasGoogle = !!(process.env.GOOGLE_API_KEY || process.env.GEMINI_API_KEY);
   const hasOpenai = !!process.env.OPENAI_API_KEY;
 
   if (hasGoogle && !hasOpenai) return "google";
@@ -265,235 +249,21 @@ function detectProvider(args: CliArgs): Provider {
   if (hasGoogle && hasOpenai) return "google";
 
   throw new Error(
-    "No API key found. Set GOOGLE_API_KEY or OPENAI_API_KEY.\n" +
+    "No API key found. Set GOOGLE_API_KEY, GEMINI_API_KEY, or OPENAI_API_KEY.\n" +
       "Create ~/.baoyu-skills/.env or <cwd>/.baoyu-skills/.env with your keys."
   );
 }
 
-function getDefaultModel(provider: Provider): string {
+type ProviderModule = {
+  getDefaultModel: () => string;
+  generateImage: (prompt: string, model: string, args: CliArgs) => Promise<Uint8Array>;
+};
+
+async function loadProviderModule(provider: Provider): Promise<ProviderModule> {
   if (provider === "google") {
-    return process.env.GOOGLE_IMAGE_MODEL || "gemini-3-pro-image-preview";
+    return (await import("./providers/google")) as ProviderModule;
   }
-  return process.env.OPENAI_IMAGE_MODEL || "gpt-image-1.5";
-}
-
-function isGoogleMultimodal(model: string): boolean {
-  return GOOGLE_MULTIMODAL_MODELS.some((m) => model.includes(m));
-}
-
-function isGoogleImagen(model: string): boolean {
-  return GOOGLE_IMAGEN_MODELS.some((m) => model.includes(m));
-}
-
-function buildPromptWithAspect(prompt: string, ar: string | null, quality: Quality): string {
-  let result = prompt;
-  if (ar) {
-    result += ` Aspect ratio: ${ar}.`;
-  }
-  if (quality === "2k") {
-    result += " High resolution 2048px.";
-  }
-  return result;
-}
-
-function parseAspectRatio(ar: string): { width: number; height: number } | null {
-  const match = ar.match(/^(\d+(?:\.\d+)?):(\d+(?:\.\d+)?)$/);
-  if (!match) return null;
-  const w = parseFloat(match[1]!);
-  const h = parseFloat(match[2]!);
-  if (w <= 0 || h <= 0) return null;
-  return { width: w, height: h };
-}
-
-function getOpenAISize(ar: string | null, quality: Quality): string {
-  const base = quality === "2k" ? 2048 : 1024;
-
-  if (!ar) return `${base}x${base}`;
-
-  const parsed = parseAspectRatio(ar);
-  if (!parsed) return `${base}x${base}`;
-
-  const ratio = parsed.width / parsed.height;
-
-  if (Math.abs(ratio - 1) < 0.1) return `${base}x${base}`;
-  if (ratio > 1.5) return quality === "2k" ? "2048x1024" : "1792x1024";
-  if (ratio < 0.67) return quality === "2k" ? "1024x2048" : "1024x1792";
-  return `${base}x${base}`;
-}
-
-async function readImageAsBase64(p: string): Promise<{ data: string; mimeType: string }> {
-  const buf = await readFile(p);
-  const ext = path.extname(p).toLowerCase();
-  let mimeType = "image/png";
-  if (ext === ".jpg" || ext === ".jpeg") mimeType = "image/jpeg";
-  else if (ext === ".gif") mimeType = "image/gif";
-  else if (ext === ".webp") mimeType = "image/webp";
-  return { data: buf.toString("base64"), mimeType };
-}
-
-async function generateWithGoogleMultimodal(
-  prompt: string,
-  model: string,
-  args: CliArgs
-): Promise<Uint8Array> {
-  const { generateText } = await import("ai");
-  const { createGoogleGenerativeAI } = await import("@ai-sdk/google");
-
-  const google = createGoogleGenerativeAI({
-    apiKey: process.env.GOOGLE_API_KEY,
-    baseURL: process.env.GOOGLE_BASE_URL,
-  });
-
-  const fullPrompt = buildPromptWithAspect(prompt, args.aspectRatio, args.quality);
-
-  const messages: any[] = [];
-  const content: any[] = [];
-
-  for (const refPath of args.referenceImages) {
-    const { data, mimeType } = await readImageAsBase64(refPath);
-    content.push({ type: "image", image: data, mimeType });
-  }
-  content.push({ type: "text", text: fullPrompt });
-
-  messages.push({ role: "user", content });
-
-  const result = await generateText({
-    model: google(model, { useSearchGrounding: false }),
-    messages,
-    providerOptions: {
-      google: {
-        responseModalities: ["TEXT", "IMAGE"],
-      },
-    },
-  });
-
-  const files = (result as any).files;
-  if (!files || files.length === 0) {
-    const expRes = (result as any).response?.body?.candidates?.[0]?.content?.parts;
-    if (expRes) {
-      for (const part of expRes) {
-        if (part.inlineData?.data) {
-          return Uint8Array.from(Buffer.from(part.inlineData.data, "base64"));
-        }
-      }
-    }
-    throw new Error("No image in response");
-  }
-
-  const img = files[0];
-  if (img.uint8Array) return img.uint8Array;
-  if (img.base64) return Uint8Array.from(Buffer.from(img.base64, "base64"));
-
-  throw new Error("Cannot extract image data");
-}
-
-async function generateWithGoogleImagen(
-  prompt: string,
-  model: string,
-  args: CliArgs
-): Promise<Uint8Array> {
-  const { experimental_generateImage: generateImage } = await import("ai");
-  const { createGoogleGenerativeAI } = await import("@ai-sdk/google");
-
-  const google = createGoogleGenerativeAI({
-    apiKey: process.env.GOOGLE_API_KEY,
-    baseURL: process.env.GOOGLE_BASE_URL,
-  });
-
-  const fullPrompt = buildPromptWithAspect(prompt, args.aspectRatio, args.quality);
-
-  const result = await generateImage({
-    model: google.image(model),
-    prompt: fullPrompt,
-    n: args.n,
-    aspectRatio: args.aspectRatio || undefined,
-  });
-
-  const img = result.images[0];
-  if (!img) throw new Error("No image in response");
-
-  if (img.uint8Array) return img.uint8Array;
-  if (img.base64) return Uint8Array.from(Buffer.from(img.base64, "base64"));
-
-  throw new Error("Cannot extract image data");
-}
-
-async function generateWithOpenAI(
-  prompt: string,
-  model: string,
-  args: CliArgs
-): Promise<Uint8Array> {
-  const baseURL = process.env.OPENAI_BASE_URL || "https://api.openai.com/v1";
-  const apiKey = process.env.OPENAI_API_KEY;
-
-  if (!apiKey) throw new Error("OPENAI_API_KEY is required");
-
-  const size = args.size || getOpenAISize(args.aspectRatio, args.quality);
-
-  const body: Record<string, any> = {
-    model,
-    prompt,
-    size,
-  };
-
-  if (model.includes("dall-e-3")) {
-    body.quality = args.quality === "2k" ? "hd" : "standard";
-  }
-
-  const res = await fetch(`${baseURL}/images/generations`, {
-    method: "POST",
-    headers: {
-      "Content-Type": "application/json",
-      Authorization: `Bearer ${apiKey}`,
-    },
-    body: JSON.stringify(body),
-  });
-
-  if (!res.ok) {
-    const err = await res.text();
-    throw new Error(`OpenAI API error: ${err}`);
-  }
-
-  const result = (await res.json()) as { data: Array<{ url?: string; b64_json?: string }> };
-  const img = result.data[0];
-
-  if (img?.b64_json) {
-    return Uint8Array.from(Buffer.from(img.b64_json, "base64"));
-  }
-
-  if (img?.url) {
-    const imgRes = await fetch(img.url);
-    if (!imgRes.ok) throw new Error("Failed to download image");
-    const buf = await imgRes.arrayBuffer();
-    return new Uint8Array(buf);
-  }
-
-  throw new Error("No image in response");
-}
-
-async function generate(
-  provider: Provider,
-  model: string,
-  prompt: string,
-  args: CliArgs
-): Promise<Uint8Array> {
-  if (provider === "google") {
-    if (isGoogleMultimodal(model)) {
-      return generateWithGoogleMultimodal(prompt, model, args);
-    }
-    if (isGoogleImagen(model)) {
-      if (args.referenceImages.length > 0) {
-        console.error("Warning: Reference images not supported with Imagen models, ignoring.");
-      }
-      return generateWithGoogleImagen(prompt, model, args);
-    }
-    return generateWithGoogleMultimodal(prompt, model, args);
-  }
-
-  if (args.referenceImages.length > 0) {
-    console.error("Warning: Reference images not supported with OpenAI, ignoring.");
-  }
-  return generateWithOpenAI(prompt, model, args);
+  return (await import("./providers/openai")) as ProviderModule;
 }
 
 async function main(): Promise<void> {
@@ -525,7 +295,8 @@ async function main(): Promise<void> {
   }
 
   const provider = detectProvider(args);
-  const model = args.model || getDefaultModel(provider);
+  const providerModule = await loadProviderModule(provider);
+  const model = args.model || providerModule.getDefaultModel();
   const outputPath = normalizeOutputImagePath(args.imagePath);
 
   let imageData: Uint8Array;
@@ -533,7 +304,7 @@ async function main(): Promise<void> {
 
   while (true) {
     try {
-      imageData = await generate(provider, model, prompt, args);
+      imageData = await providerModule.generateImage(prompt, model, args);
       break;
     } catch (e) {
       if (!retried) {
diff --git a/skills/baoyu-image-gen/scripts/providers/google.ts b/skills/baoyu-image-gen/scripts/providers/google.ts
new file mode 100644
index 0000000..4b3ef86
--- /dev/null
+++ b/skills/baoyu-image-gen/scripts/providers/google.ts
@@ -0,0 +1,149 @@
+import path from "node:path";
+import { readFile } from "node:fs/promises";
+import type { CliArgs } from "../types";
+
+const GOOGLE_MULTIMODAL_MODELS = ["gemini-3-pro-image-preview"];
+const GOOGLE_IMAGEN_MODELS = ["imagen-3.0-generate-002", "imagen-3.0-generate-001"];
+
+export function getDefaultModel(): string {
+  return process.env.GOOGLE_IMAGE_MODEL || "gemini-3-pro-image-preview";
+}
+
+function isGoogleMultimodal(model: string): boolean {
+  return GOOGLE_MULTIMODAL_MODELS.some((m) => model.includes(m));
+}
+
+function isGoogleImagen(model: string): boolean {
+  return GOOGLE_IMAGEN_MODELS.some((m) => model.includes(m));
+}
+
+function getGoogleApiKey(): string | null {
+  return process.env.GOOGLE_API_KEY || process.env.GEMINI_API_KEY || null;
+}
+
+function getGoogleImageSize(args: CliArgs): "1K" | "2K" | "4K" {
+  if (args.imageSize) return args.imageSize as "1K" | "2K" | "4K";
+  return args.quality === "2k" ? "2K" : "1K";
+}
+
+function buildPromptWithAspect(prompt: string, ar: string | null, quality: CliArgs["quality"]): string {
+  let result = prompt;
+  if (ar) {
+    result += ` Aspect ratio: ${ar}.`;
+  }
+  if (quality === "2k") {
+    result += " High resolution 2048px.";
+  }
+  return result;
+}
+
+async function readImageAsBase64(p: string): Promise<{ data: string; mimeType: string }> {
+  const buf = await readFile(p);
+  const ext = path.extname(p).toLowerCase();
+  let mimeType = "image/png";
+  if (ext === ".jpg" || ext === ".jpeg") mimeType = "image/jpeg";
+  else if (ext === ".gif") mimeType = "image/gif";
+  else if (ext === ".webp") mimeType = "image/webp";
+  return { data: buf.toString("base64"), mimeType };
+}
+
+async function generateWithGemini(
+  prompt: string,
+  model: string,
+  args: CliArgs
+): Promise<Uint8Array> {
+  const { GoogleGenAI } = await import("@google/genai");
+
+  const apiKey = getGoogleApiKey();
+  if (!apiKey) throw new Error("GOOGLE_API_KEY or GEMINI_API_KEY is required");
+
+  const ai = new GoogleGenAI({
+    apiKey,
+    httpOptions: {
+      baseUrl: process.env.GOOGLE_BASE_URL || undefined,
+    },
+  });
+
+  const input: Array<{ type: "text" | "image"; text?: string; data?: string; mime_type?: string }> = [];
+  for (const refPath of args.referenceImages) {
+    const { data, mimeType } = await readImageAsBase64(refPath);
+    input.push({ type: "image", data, mime_type: mimeType });
+  }
+  input.push({ type: "text", text: prompt });
+
+  const imageConfig: { image_size: "1K" | "2K" | "4K"; aspect_ratio?: string } = {
+    image_size: getGoogleImageSize(args),
+  };
+  if (args.aspectRatio) {
+    imageConfig.aspect_ratio = args.aspectRatio;
+  }
+
+  console.log("Generating image with Gemini...", imageConfig);
+  const interaction = await ai.interactions.create({
+    model,
+    input,
+    response_modalities: ["image"],
+    generation_config: {
+      image_config: imageConfig,
+    },
+  });
+  console.log("Generation completed.");
+
+  for (const output of interaction.outputs || []) {
+    if (output.type === "image" && output.data) {
+      return Uint8Array.from(Buffer.from(output.data, "base64"));
+    }
+  }
+
+  throw new Error("No image in response");
+}
+
+async function generateWithImagen(
+  prompt: string,
+  model: string,
+  args: CliArgs
+): Promise<Uint8Array> {
+  const { experimental_generateImage: generateImage } = await import("ai");
+  const { createGoogleGenerativeAI } = await import("@ai-sdk/google");
+
+  const google = createGoogleGenerativeAI({
+    apiKey: getGoogleApiKey() || undefined,
+    baseURL: process.env.GOOGLE_BASE_URL,
+  });
+
+  const fullPrompt = buildPromptWithAspect(prompt, args.aspectRatio, args.quality);
+
+  const result = await generateImage({
+    model: google.image(model),
+    prompt: fullPrompt,
+    n: args.n,
+    aspectRatio: args.aspectRatio || undefined,
+  });
+
+  const img = result.images[0];
+  if (!img) throw new Error("No image in response");
+
+  if (img.uint8Array) return img.uint8Array;
+  if (img.base64) return Uint8Array.from(Buffer.from(img.base64, "base64"));
+
+  throw new Error("Cannot extract image data");
+}
+
+export async function generateImage(
+  prompt: string,
+  model: string,
+  args: CliArgs
+): Promise<Uint8Array> {
+  if (isGoogleImagen(model)) {
+    if (args.referenceImages.length > 0) {
+      console.error("Warning: Reference images not supported with Imagen models, ignoring.");
+    }
+    return generateWithImagen(prompt, model, args);
+  }
+
+  if (!isGoogleMultimodal(model) && args.referenceImages.length > 0) {
+    console.error("Warning: Reference images are only supported with Gemini multimodal models.");
+  }
+
+  return generateWithGemini(prompt, model, args);
+}
diff --git a/skills/baoyu-image-gen/scripts/providers/openai.ts b/skills/baoyu-image-gen/scripts/providers/openai.ts
new file mode 100644
index 0000000..f42a762
--- /dev/null
+++ b/skills/baoyu-image-gen/scripts/providers/openai.ts
@@ -0,0 +1,114 @@
+import type { CliArgs } from "../types";
+
+export function getDefaultModel(): string {
+  return process.env.OPENAI_IMAGE_MODEL || "gpt-image-1.5";
+}
+
+function parseAspectRatio(ar: string): { width: number; height: number } | null {
+  const match = ar.match(/^(\d+(?:\.\d+)?):(\d+(?:\.\d+)?)$/);
+  if (!match) return null;
+  const w = parseFloat(match[1]!);
+  const h = parseFloat(match[2]!);
+  if (w <= 0 || h <= 0) return null;
+  return { width: w, height: h };
+}
+
+type SizeMapping = {
+  square: string;
+  landscape: string;
+  portrait: string;
+};
+
+function getOpenAISize(
+  model: string,
+  ar: string | null,
+  quality: CliArgs["quality"]
+): string {
+  const isDalle3 = model.includes("dall-e-3");
+  const isDalle2 = model.includes("dall-e-2");
+
+  if (isDalle2) {
+    return "1024x1024";
+  }
+
+  const sizes: SizeMapping = isDalle3
+    ? {
+        square: "1024x1024",
+        landscape: "1792x1024",
+        portrait: "1024x1792",
+      }
+    : {
+        square: "1024x1024",
+        landscape: "1536x1024",
+        portrait: "1024x1536",
+      };
+
+  if (!ar) return sizes.square;
+
+  const parsed = parseAspectRatio(ar);
+  if (!parsed) return sizes.square;
+
+  const ratio = parsed.width / parsed.height;
+
+  if (Math.abs(ratio - 1) < 0.1) return sizes.square;
+  if (ratio > 1.5) return sizes.landscape;
+  if (ratio < 0.67) return sizes.portrait;
+  return sizes.square;
+}
+
+export async function generateImage(
+  prompt: string,
+  model: string,
+  args: CliArgs
+): Promise<Uint8Array> {
+  const baseURL = process.env.OPENAI_BASE_URL || "https://api.openai.com/v1";
+  const apiKey = process.env.OPENAI_API_KEY;
+
+  if (!apiKey) throw new Error("OPENAI_API_KEY is required");
+
+  if (args.referenceImages.length > 0) {
+    console.error("Warning: Reference images not supported with OpenAI, ignoring.");
+  }
+
+  const size = args.size || getOpenAISize(model, args.aspectRatio, args.quality);
+
+  const body: Record<string, any> = {
+    model,
+    prompt,
+    size,
+  };
+
+  if (model.includes("dall-e-3")) {
+    body.quality = args.quality === "2k" ? "hd" : "standard";
+  }
+
+  const res = await fetch(`${baseURL}/images/generations`, {
+    method: "POST",
+    headers: {
+      "Content-Type": "application/json",
+      Authorization: `Bearer ${apiKey}`,
+    },
+    body: JSON.stringify(body),
+  });
+
+  if (!res.ok) {
+    const err = await res.text();
+    throw new Error(`OpenAI API error: ${err}`);
+  }
+
+  const result = (await res.json()) as { data: Array<{ url?: string; b64_json?: string }> };
+  const img = result.data[0];
+
+  if (img?.b64_json) {
+    return Uint8Array.from(Buffer.from(img.b64_json, "base64"));
+  }
+
+  if (img?.url) {
+    const imgRes = await fetch(img.url);
+    if (!imgRes.ok) throw new Error("Failed to download image");
+    const buf = await imgRes.arrayBuffer();
+    return new Uint8Array(buf);
+  }
+
+  throw new Error("No image in response");
+}
diff --git a/skills/baoyu-image-gen/scripts/types.ts b/skills/baoyu-image-gen/scripts/types.ts
new file mode 100644
index 0000000..6121391
--- /dev/null
+++ b/skills/baoyu-image-gen/scripts/types.ts
@@ -0,0 +1,18 @@
+export type Provider = "google" | "openai";
+export type Quality = "normal" | "2k";
+
+export type CliArgs = {
+  prompt: string | null;
+  promptFiles: string[];
+  imagePath: string | null;
+  provider: Provider | null;
+  model: string | null;
+  aspectRatio: string | null;
+  size: string | null;
+  quality: Quality;
+  imageSize: string | null;
+  referenceImages: string[];
+  n: number;
+  json: boolean;
+  help: boolean;
+};