chore: release v1.113.0

feat(baoyu-imagine): add DashScope Wan 2.7 image model support (#141 )
* feat(baoyu-imagine): add DashScope Wan 2.7 image model support Closes #139. Adds the new `wan2.7-image-pro` and `wan2.7-image` model family to the DashScope provider so users can call Wan 2.7 directly through the official Aliyun (Bailian) API instead of going through Replicate. - Register `wan2.7-image-pro` and `wan2.7-image` as a new `wan27` family in the DashScope provider with their own size resolution rules: pixel range `[768*768, 4096*4096]` for `wan2.7-image-pro` text-to-image, `[768*768, 2048*2048]` for `wan2.7-image-pro` with refs and for the base `wan2.7-image` model in any mode, with aspect ratios validated against the documented `[1:8, 8:1]` band. - Allow up to 9 reference images per request (image editing / multi-image fusion). Local files are inlined as base64 data URLs; `http(s)://` paths are forwarded as-is. Other DashScope models still reject `--ref` with a hint to switch to a wan2.7 model or another provider. - Drop `prompt_extend` from the request body for the Wan 2.7 family (not part of the Wan 2.7 API surface) and skip the Qwen-only negative prompt for this family. - Allow `--provider dashscope --ref ...` in `detectProvider` so users can opt into Wan 2.7 reference workflows, while keeping Wan 2.7 out of the auto-detect ref priority list. - Add provider, reference, and usage-example documentation, plus unit tests covering family routing, size derivation across the three pixel-budget modes, ratio rejection, explicit-size validation, and the new `--provider dashscope` ref opt-in path. Made-with: Cursor * fix(baoyu-imagine): force n=1 for DashScope wan2.7 to avoid silent multi-image billing Cross-checked the implementation against the official Wan 2.7 image generation & editing API reference and found that the API defaults `parameters.n` to 4 in non-collage mode (1-4 range, billed per image). baoyu-imagine has single-image save semantics — only the first image in the response is kept — so without an explicit `n: 1` users would silently pay for 3 discarded images per request. - Always send `parameters.n: 1` in the wan2.7 request body - Reject `--n > 1` for wan2.7 with a clear error pointing at the single-image save semantics - Add tests asserting the request body shape (n=1, no prompt_extend, no negative_prompt) and the --n>1 rejection - Document the defaults-vs-skill mismatch in the dashscope reference Made-with: Cursor * Fix DashScope Wan 2.7 review feedback
2026-07-12 13:59:47 +08:00 · 2026-04-25 15:03:22 -05:00 · 2026-04-25 14:54:08 -05:00 · 2026-04-24 02:15:57 -05:00 · 2026-04-24 02:15:30 -05:00 · 2026-04-24 02:15:25 -05:00
45 changed files with 1344 additions and 218 deletions
@@ -6,7 +6,7 @@
  },
  "metadata": {
    "description": "Skills shared by Baoyu for improving daily work efficiency",
-    "version": "1.109.0"
+    "version": "1.113.0"
  },
  "plugins": [
    {
@@ -2,6 +2,40 @@

 English | [中文](./CHANGELOG.zh.md)

+## 1.113.0 - 2026-04-25
+
+### Features
+- `baoyu-imagine`: add DashScope Wan 2.7 image model support (`wan2.7-image-pro` and `wan2.7-image`) directly through the official Aliyun (Bailian) API. Supports text-to-image, image editing, and multi-image fusion with up to 9 reference images, with documented `[1:8, 8:1]` aspect ratio validation and per-mode pixel-budget rules. Forces `parameters.n: 1` to match baoyu-imagine's single-image save semantics and explicitly rejects `--n > 1` to prevent silent multi-image billing (the API defaults to `n=4` in non-collage mode). Allows `--provider dashscope --ref ...` opt-in for Wan 2.7 reference workflows.
+
+## 1.112.0 - 2026-04-24
+
+### Features
+- `baoyu-article-illustrator`: make `hand-drawn-edu` (infographic + sketch-notes + macaron) the universal fallback preset when content analysis surfaces no strong signal — warm cream paper, black hand-drawn lines, soft pastel section blocks. Elevate `sketch-notes` to primary style across infographic / flowchart / comparison / framework auto-selection; rewrite the sketch-notes style spec (macaron palette, canonical single-page layout, diagram-only rule); add matching prompt block and workflow defaults.
+- `baoyu-article-illustrator`: add `hand-drawn-edu-flow` (flowchart) and `hand-drawn-edu-compare` (comparison) presets for the same warm educational style.
+
+### Breaking Changes
+- `baoyu-article-illustrator`: `hand-drawn-edu` preset now maps to `infographic` instead of `flowchart`. Users relying on the previous flowchart behavior should switch to the new `hand-drawn-edu-flow` preset.
+
+### Fixes
+- `baoyu-post-to-x`: add entry point guard to `scripts/md-to-html.ts` so that importing `parseMarkdown` from `x-article.ts` no longer triggers the CLI entry point. Mirrors the same fix applied to `baoyu-post-to-weibo`.
+
+## 1.111.1 - 2026-04-21
+
+### Documentation
+- Add a top-level `## Confirmation Policy` section to every image-generating skill (`baoyu-infographic`, `baoyu-cover-image`, `baoyu-slide-deck`, `baoyu-image-cards`, `baoyu-xhs-images`, `baoyu-article-illustrator`) as a single source of truth: explicit skill invocation, keyword shortcuts, EXTEND.md defaults, and auto-recommendations are recommendation inputs only — they never authorize skipping the confirmation step. Opt-out requires an explicit current-request signal (`--no-confirm` / `--quick` / `--yes` / "直接生成" / equivalent).
+- `baoyu-infographic`: consolidate the scattered reminders (previously repeated across Step 5, Step 6, Default combination, Keyword Shortcuts, and the preferences docs) into a single policy section referenced from Step 4's hard gate.
+
+## 1.111.0 - 2026-04-21
+
+### Refactor
+- Unify image-backend resolution across all image-consuming skills (`baoyu-infographic`, `baoyu-comic`, `baoyu-cover-image`, `baoyu-image-cards`, `baoyu-article-illustrator`, `baoyu-slide-deck`, `baoyu-xhs-images`): add a single `preferred_image_backend` preference field (`auto | ask | <backend-id>`) and replace the stateless ask-once rule with a 4-step resolution (current-request override → saved preference → auto-select → ask). Runtime-native tools (Codex `imagegen`, Hermes `image_generate`) are preferred by default; existing `EXTEND.md` files without the field are treated as `auto` with no schema bump.
+- Add a top-level `## Changing Preferences` section to each image-consuming skill as a first-class surface for pinning the backend and editing common one-line preferences.
+
+## 1.110.0 - 2026-04-21
+
+### Features
+- `baoyu-imagine`: add `gpt-image-2` support for OpenAI image generation and edits, make it the default OpenAI model, and document the official size/quality mapping, custom-size constraints, and Azure deployment guidance
+
 ## 1.109.0 - 2026-04-21

 ### Features
@@ -2,6 +2,40 @@

 [English](./CHANGELOG.md) | 中文

+## 1.113.0 - 2026-04-25
+
+### 新功能
+- `baoyu-imagine`：新增 DashScope Wan 2.7 图像模型支持（`wan2.7-image-pro` 与 `wan2.7-image`），通过阿里云百炼官方 API 直接调用，无需经 Replicate 转发。支持文生图、图像编辑、多图融合（最多 9 张参考图），按官方文档校验 `[1:8, 8:1]` 宽高比范围，并按模式应用不同的像素预算规则。强制 `parameters.n: 1` 以匹配 baoyu-imagine 的单图保存语义，显式拒绝 `--n > 1`，避免在用户不知情的情况下产生多图计费（API 在非拼图模式下默认 `n=4`）。允许通过 `--provider dashscope --ref ...` 显式启用 Wan 2.7 参考图工作流。
+
+## 1.112.0 - 2026-04-24
+
+### 新功能
+- `baoyu-article-illustrator`：当内容分析未检测到明确信号时，将 `hand-drawn-edu`（infographic + sketch-notes + macaron）作为通用默认预设 —— 暖奶油色纸面背景、黑色手绘线条、柔和马卡龙色块。`sketch-notes` 升级为 infographic / flowchart / comparison / framework 自动选择的首选风格；重写 sketch-notes 风格规范（马卡龙调色板、标准单页布局、仅限示意图的规则）；新增对应的 prompt 模板块和默认工作流规则。
+- `baoyu-article-illustrator`：新增 `hand-drawn-edu-flow`（flowchart）和 `hand-drawn-edu-compare`（comparison）两个预设，保持相同的温暖教育风格。
+
+### 破坏性变更
+- `baoyu-article-illustrator`：`hand-drawn-edu` 预设的类型由 `flowchart` 改为 `infographic`。依赖原有流程图行为的用户请改用新增的 `hand-drawn-edu-flow` 预设。
+
+### 修复
+- `baoyu-post-to-x`：为 `scripts/md-to-html.ts` 添加入口守卫，确保 `x-article.ts` 导入 `parseMarkdown` 时不再触发 CLI 入口逻辑。与 `baoyu-post-to-weibo` 此前的修复保持一致。
+
+## 1.111.1 - 2026-04-21
+
+### 文档
+- 为每个图片生成类技能（`baoyu-infographic`、`baoyu-cover-image`、`baoyu-slide-deck`、`baoyu-image-cards`、`baoyu-xhs-images`、`baoyu-article-illustrator`）新增顶级 `## Confirmation Policy` 章节作为单一事实源：显式调用技能、关键词快捷方式、EXTEND.md 偏好、自动推荐都只是"推荐输入"，不授权跳过确认步骤。跳过确认必须由当前请求中的明确信号触发（`--no-confirm` / `--quick` / `--yes` / "直接生成" / 同义表达）。
+- `baoyu-infographic`：将原先散落在 Step 5、Step 6、Default combination、Keyword Shortcuts 及 preferences 文档中的重复提醒合并为单一策略章节，由 Step 4 的 hard gate 引用。
+
+## 1.111.0 - 2026-04-21
+
+### 重构
+- 统一所有图片生成类技能（`baoyu-infographic`、`baoyu-comic`、`baoyu-cover-image`、`baoyu-image-cards`、`baoyu-article-illustrator`、`baoyu-slide-deck`、`baoyu-xhs-images`）的后端选择规则：新增单一 `preferred_image_backend` 偏好字段（`auto | ask | <backend-id>`），用 4 步解析规则（当前请求覆盖 → 已保存偏好 → 自动选择 → 询问用户）替换原有的无状态询问规则。默认优先使用运行时原生工具（如 Codex `imagegen`、Hermes `image_generate`）；未设置该字段的现有 `EXTEND.md` 文件视为 `auto`，无需升级 schema 版本。
+- 在每个图片技能中新增顶级 `## Changing Preferences` 章节,作为固定后端和修改常用偏好的一级入口。
+
+## 1.110.0 - 2026-04-21
+
+### 新功能
+- `baoyu-imagine`：新增 `gpt-image-2` 支持，用于 OpenAI 图像生成与编辑；将其设为默认 OpenAI 模型，并补齐官方尺寸/质量映射、自定义尺寸约束与 Azure 部署说明
+
 ## 1.109.0 - 2026-04-21

 ### 新功能
@@ -714,7 +714,7 @@ AI-powered generation backends.

 #### baoyu-imagine

-AI SDK-based image generation using OpenAI, Azure OpenAI, Google, OpenRouter, DashScope (Aliyun Tongyi Wanxiang), MiniMax, Jimeng (即梦), Seedream (豆包), and Replicate APIs. Supports text-to-image, reference images, aspect ratios, custom sizes, batch generation, and quality presets.
+AI SDK-based image generation using OpenAI GPT Image 2, Azure OpenAI, Google, OpenRouter, DashScope (Aliyun Tongyi Wanxiang), MiniMax, Jimeng (即梦), Seedream (豆包), and Replicate APIs. Supports text-to-image, reference images, aspect ratios, custom sizes, batch generation, and quality presets.

 ```bash
 # Basic generation (auto-detect provider)
@@ -727,10 +727,10 @@ AI SDK-based image generation using OpenAI, Azure OpenAI, Google, OpenRouter, Da
 /baoyu-imagine --prompt "A banner" --image banner.png --quality 2k

 # Specific provider
-/baoyu-imagine --prompt "A cat" --image cat.png --provider openai
+/baoyu-imagine --prompt "A cat" --image cat.png --provider openai --model gpt-image-2

 # Azure OpenAI (model = deployment name)
-/baoyu-imagine --prompt "A cat" --image cat.png --provider azure --model gpt-image-1.5
+/baoyu-imagine --prompt "A cat" --image cat.png --provider azure --model gpt-image-2

 # OpenRouter
 /baoyu-imagine --prompt "A cat" --image cat.png --provider openrouter
@@ -786,7 +786,7 @@ AI SDK-based image generation using OpenAI, Azure OpenAI, Google, OpenRouter, Da
 | `--provider` | `google`, `openai`, `azure`, `openrouter`, `dashscope`, `zai`, `minimax`, `jimeng`, `seedream`, or `replicate` |
 | `--model`, `-m` | Model ID or deployment name. Azure uses deployment name; OpenRouter uses full model IDs; Z.AI uses `glm-image`; MiniMax uses `image-01` / `image-01-live` |
 | `--ar` | Aspect ratio (e.g., `16:9`, `1:1`, `4:3`) |
-| `--size` | Size (e.g., `1024x1024`) |
+| `--size` | Size (e.g., `1024x1024`; `gpt-image-2` accepts valid custom sizes up to 3840px max edge) |
 | `--quality` | `normal` or `2k` (default: `2k`) |
 | `--imageSize` | `1K`, `2K`, or `4K` for Google/OpenRouter |
 | `--imageApiDialect` | `openai-native` or `ratio-metadata` for OpenAI-compatible gateways |
@@ -810,9 +810,9 @@ AI SDK-based image generation using OpenAI, Azure OpenAI, Google, OpenRouter, Da
 | `JIMENG_ACCESS_KEY_ID` | Jimeng Volcengine access key | - |
 | `JIMENG_SECRET_ACCESS_KEY` | Jimeng Volcengine secret key | - |
 | `ARK_API_KEY` | Seedream Volcengine ARK API key | - |
-| `OPENAI_IMAGE_MODEL` | OpenAI model | `gpt-image-1.5` |
+| `OPENAI_IMAGE_MODEL` | OpenAI model | `gpt-image-2` |
 | `AZURE_OPENAI_DEPLOYMENT` | Azure default deployment name | - |
-| `AZURE_OPENAI_IMAGE_MODEL` | Backward-compatible Azure deployment/model alias | `gpt-image-1.5` |
+| `AZURE_OPENAI_IMAGE_MODEL` | Backward-compatible Azure deployment/model alias | `gpt-image-2` |
 | `OPENROUTER_IMAGE_MODEL` | OpenRouter model | `google/gemini-3.1-flash-image-preview` |
 | `GOOGLE_IMAGE_MODEL` | Google model | `gemini-3-pro-image-preview` |
 | `DASHSCOPE_IMAGE_MODEL` | DashScope model | `qwen-image-2.0-pro` |
@@ -1132,14 +1132,14 @@ mkdir -p ~/.baoyu-skills
 cat > ~/.baoyu-skills/.env << 'EOF'
 # OpenAI
 OPENAI_API_KEY=sk-xxx
-OPENAI_IMAGE_MODEL=gpt-image-1.5
+OPENAI_IMAGE_MODEL=gpt-image-2
 # OPENAI_BASE_URL=https://api.openai.com/v1
 # OPENAI_IMAGE_USE_CHAT=false

 # Azure OpenAI
 AZURE_OPENAI_API_KEY=xxx
 AZURE_OPENAI_BASE_URL=https://your-resource.openai.azure.com
-AZURE_OPENAI_DEPLOYMENT=gpt-image-1.5
+AZURE_OPENAI_DEPLOYMENT=gpt-image-2
 # AZURE_API_VERSION=2025-04-01-preview

 # OpenRouter
@@ -714,7 +714,7 @@ AI 驱动的生成后端。

 #### baoyu-imagine

-基于 AI SDK 的图像生成，支持 OpenAI、Azure OpenAI、Google、OpenRouter、DashScope（阿里通义万相）、MiniMax、即梦（Jimeng）、豆包（Seedream）和 Replicate API。支持文生图、参考图、宽高比、自定义尺寸、批量生成和质量预设。
+基于 AI SDK 的图像生成，支持 OpenAI GPT Image 2、Azure OpenAI、Google、OpenRouter、DashScope（阿里通义万相）、MiniMax、即梦（Jimeng）、豆包（Seedream）和 Replicate API。支持文生图、参考图、宽高比、自定义尺寸、批量生成和质量预设。

 ```bash
 # 基础生成（自动检测服务商）
@@ -727,10 +727,10 @@ AI 驱动的生成后端。
 /baoyu-imagine --prompt "横幅图" --image banner.png --quality 2k

 # 指定服务商
-/baoyu-imagine --prompt "一只猫" --image cat.png --provider openai
+/baoyu-imagine --prompt "一只猫" --image cat.png --provider openai --model gpt-image-2

 # Azure OpenAI（model 为部署名称）
-/baoyu-imagine --prompt "一只猫" --image cat.png --provider azure --model gpt-image-1.5
+/baoyu-imagine --prompt "一只猫" --image cat.png --provider azure --model gpt-image-2

 # OpenRouter
 /baoyu-imagine --prompt "一只猫" --image cat.png --provider openrouter
@@ -786,7 +786,7 @@ AI 驱动的生成后端。
 | `--provider` | `google`、`openai`、`azure`、`openrouter`、`dashscope`、`zai`、`minimax`、`jimeng`、`seedream` 或 `replicate` |
 | `--model`, `-m` | 模型 ID 或部署名。Azure 使用部署名；OpenRouter 使用完整模型 ID；Z.AI 使用 `glm-image`；MiniMax 使用 `image-01` / `image-01-live` |
 | `--ar` | 宽高比（如 `16:9`、`1:1`、`4:3`） |
-| `--size` | 尺寸（如 `1024x1024`） |
+| `--size` | 尺寸（如 `1024x1024`；`gpt-image-2` 支持最长边不超过 3840px 的有效自定义尺寸） |
 | `--quality` | `normal` 或 `2k`（默认：`2k`） |
 | `--imageSize` | Google/OpenRouter 使用的 `1K`、`2K`、`4K` |
 | `--imageApiDialect` | OpenAI 兼容网关的图像 API 方言（`openai-native` 或 `ratio-metadata`） |
@@ -810,9 +810,9 @@ AI 驱动的生成后端。
 | `JIMENG_ACCESS_KEY_ID` | 即梦火山引擎 Access Key | - |
 | `JIMENG_SECRET_ACCESS_KEY` | 即梦火山引擎 Secret Key | - |
 | `ARK_API_KEY` | 豆包火山引擎 ARK API 密钥 | - |
-| `OPENAI_IMAGE_MODEL` | OpenAI 模型 | `gpt-image-1.5` |
+| `OPENAI_IMAGE_MODEL` | OpenAI 模型 | `gpt-image-2` |
 | `AZURE_OPENAI_DEPLOYMENT` | Azure 默认部署名 | - |
-| `AZURE_OPENAI_IMAGE_MODEL` | 兼容旧配置的 Azure 部署/模型别名 | `gpt-image-1.5` |
+| `AZURE_OPENAI_IMAGE_MODEL` | 兼容旧配置的 Azure 部署/模型别名 | `gpt-image-2` |
 | `OPENROUTER_IMAGE_MODEL` | OpenRouter 模型 | `google/gemini-3.1-flash-image-preview` |
 | `GOOGLE_IMAGE_MODEL` | Google 模型 | `gemini-3-pro-image-preview` |
 | `DASHSCOPE_IMAGE_MODEL` | DashScope 模型 | `qwen-image-2.0-pro` |
@@ -1132,14 +1132,14 @@ mkdir -p ~/.baoyu-skills
 cat > ~/.baoyu-skills/.env << 'EOF'
 # OpenAI
 OPENAI_API_KEY=sk-xxx
-OPENAI_IMAGE_MODEL=gpt-image-1.5
+OPENAI_IMAGE_MODEL=gpt-image-2
 # OPENAI_BASE_URL=https://api.openai.com/v1
 # OPENAI_IMAGE_USE_CHAT=false

 # Azure OpenAI
 AZURE_OPENAI_API_KEY=xxx
 AZURE_OPENAI_BASE_URL=https://your-resource.openai.azure.com
-AZURE_OPENAI_DEPLOYMENT=gpt-image-1.5
+AZURE_OPENAI_DEPLOYMENT=gpt-image-2
 # AZURE_API_VERSION=2025-04-01-preview

 # OpenRouter
@@ -4,13 +4,29 @@ Skills in this repo are loaded by multiple agent runtimes (Claude Code, Codex, H

 ## The Rule

-When a skill needs to render an image:
+When a skill needs to render an image, resolve the backend in this order:

- **Use whatever image-generation tool or skill is available** in the current runtime — e.g., Codex `imagegen`, Hermes `image_generate`, `baoyu-imagine`, or any equivalent the user has installed.
- **If multiple are available**, ask the user **once** at the start which to use (batch with any other initial questions).
- **If none are available**, tell the user and ask how to proceed.
+1. **Current-request override** — if the user names a specific backend in the current message, use it.
+2. **Saved preference** — if the skill's `EXTEND.md` sets `preferred_image_backend` to a backend available right now, use it.
+3. **Auto-select** (when the preference is `auto`, unset, or the pinned backend isn't available):
+   - If the current runtime exposes a native image tool (e.g., Codex `imagegen`, Hermes `image_generate`), use it. Runtime-native tools are preferred by default — agents that know their own tool inventory should surface the native one here.
+   - Otherwise, if exactly one non-native backend is installed (e.g., `baoyu-imagine`), use it.
+   - Otherwise (multiple non-native backends with no runtime-native tool), ask the user once — batch with any other initial questions.
+4. **If none are available**, tell the user and ask how to proceed.

-No explicit priority between runtime-native tools and repo skills — treat them equivalently and let the user decide when there's a choice. No persisted preference mechanism; the question is cheap, and the rule is stateless.
+Setting `preferred_image_backend: ask` forces the step-3 prompt every run regardless of available backends.
+
+## The Preference Field
+
+Each image-consuming skill's `EXTEND.md` carries a single `preferred_image_backend` field:
+
+| Value | Meaning |
+|---|---|
+| `auto` (default) | Apply the auto-select rule — runtime-native preferred, fall back to only installed backend, ask if multiple non-native. |
+| `ask` | Always confirm the backend on every run, even when a runtime-native tool exists. |
+| `<backend-id>` (e.g., `codex-imagegen`, `baoyu-imagine`, `image_generate`) | Pin this backend when available; fall back to `auto` if it isn't. |
+
+The field is **absent-equals-auto**: older `EXTEND.md` files without this field behave exactly as if `preferred_image_backend: auto` were set. No schema version bump is needed to introduce it.

 ## Prompt File Requirement (hard)

@@ -20,6 +36,8 @@ Regardless of which backend is chosen, every skill that renders images MUST writ

 Each `SKILL.md` that renders images includes **exactly one** `## Image Generation Tools` section (near the top, after `## User Input Tools` and before the main workflow) that **inlines** this rule. Skills are self-contained and cannot link to `docs/` — each skill folder must ship the rule inside its own `SKILL.md`. See [CLAUDE.md → Skill Self-Containment](../CLAUDE.md).

+Each skill's `references/config/preferences-schema.md` (and its `EXTEND.md` template in `first-time-setup.md`) lists `preferred_image_backend` alongside other preference fields. First-time setup does NOT ask the user about the backend — `auto` is set silently. Users who want to pin a specific backend edit `EXTEND.md` later, and each skill's `## Changing Preferences` section documents the common one-line edits.
+
 Concrete tool names (`imagegen`, `image_generate`, `baoyu-imagine`) in this document and in SKILL.md are **examples** — agents in other runtimes apply the rule above and substitute the local equivalent. Skill-specific parameters for these backends are illustrative; runtimes without those knobs can omit them.

 ## Backend Skills Are Exempt
@@ -1,7 +1,7 @@
 ---
 name: baoyu-article-illustrator
 description: Analyzes article structure, identifies positions requiring visual aids, generates illustrations with Type × Style × Palette three-dimension approach. Use when user asks to "illustrate article", "add images", "generate images for article", or "为文章配图".
-version: 1.57.0
+version: 1.58.0
 metadata:
  openclaw:
    homepage: https://github.com/JimLiu/baoyu-skills#baoyu-article-illustrator
@@ -23,16 +23,31 @@ Concrete `AskUserQuestion` references below are examples — substitute the loca

 ## Image Generation Tools

-When this skill needs to render an image:
+When this skill needs to render an image, resolve the backend in this order:

- **Use whatever image-generation tool or skill is available** in the current runtime — e.g., Codex `imagegen`, Hermes `image_generate`, `baoyu-imagine`, or any equivalent the user has installed.
- **If multiple are available**, ask the user **once** at the start which to use (batch with any other initial questions).
- **If none are available**, tell the user and ask how to proceed.
+1. **Current-request override** — if the user names a specific backend in the current message, use it.
+2. **Saved preference** — if `EXTEND.md` sets `preferred_image_backend` to a backend available right now, use it.
+3. **Auto-select** (when the preference is `auto`, unset, or the pinned backend isn't available):
+   - If the current runtime exposes a native image tool (e.g., Codex `imagegen`, Hermes `image_generate`), use it. Runtime-native tools are preferred by default — agents that know their own tool inventory should surface the native one here.
+   - Otherwise, if exactly one non-native backend is installed (e.g., `baoyu-imagine`), use it.
+   - Otherwise (multiple non-native backends with no runtime-native tool), ask the user once — batch with any other initial questions.
+4. **If none are available**, tell the user and ask how to proceed.
+
+Setting `preferred_image_backend: ask` forces the step-3 prompt every run regardless of available backends. Users change the pinned backend via the `## Changing Preferences` section below.

 **Prompt file requirement (hard)**: write each image's full, final prompt to a standalone file under `prompts/` (naming: `NN-{type}-[slug].md`) BEFORE invoking any backend. The backend receives the prompt file (or its content); the file is the reproducibility record and lets you switch backends without regenerating prompts.

 Concrete tool names (`imagegen`, `image_generate`, `baoyu-imagine`) above are examples — substitute the local equivalents under the same rule.

+## Confirmation Policy
+
+Default behavior: **confirm before generation**.
+
+- Treat explicit skill invocation, a file path, matched signals/presets, and `EXTEND.md` defaults as **recommendation inputs only**. None of them authorizes skipping confirmation.
+- Do **not** start Step 4 or later until the user completes Step 3.
+- Skip confirmation only when the current request explicitly says to do so, for example: "直接生成", "不用确认", "跳过确认", "按默认出图", or equivalent wording.
+- If confirmation is skipped explicitly, state the assumed type / density / style / palette / language / backend in the next user-facing update before generating.
+
 ## Reference Images

 Users may supply reference images via `--ref <files...>` or by providing file paths / pasting images in conversation. Refs guide style, palette, composition, or subject for specific illustrations.
@@ -111,6 +126,8 @@ Full procedures: [references/workflow.md](references/workflow.md#step-2-setup--a

 ### Step 3: Confirm Settings ⚠️

+**Hard gate**: this step is mandatory per the [Confirmation Policy](#confirmation-policy) — Steps 4+ cannot start until the user confirms here (or explicitly opts out with "直接生成" / equivalent wording in the current request).
+
 **ONE AskUserQuestion, max 4 Qs. Q1-Q2 REQUIRED. Q3 required unless preset chosen.**

 | Q | Options |
@@ -207,3 +224,17 @@ When input is **pasted content** (no file path), always uses `illustrations/{top
 | [references/style-presets.md](references/style-presets.md) | Preset shortcuts (type + style + palette) |
 | [references/prompt-construction.md](references/prompt-construction.md) | Prompt templates |
 | [references/config/first-time-setup.md](references/config/first-time-setup.md) | First-time setup |
+
+## Changing Preferences
+
+EXTEND.md lives at the first matching path listed in Step 1.5. Three ways to change it:
+
+- **Edit directly** — open EXTEND.md and change fields. Full schema: `references/config/preferences-schema.md`.
+- **Reconfigure interactively** — delete EXTEND.md (or ask "reconfigure baoyu-article-illustrator preferences" / "重新配置"). The next run re-triggers first-time setup.
+- **Common one-line edits**:
+  - `preferred_image_backend: auto` — default; runtime-native tool wins, falls back to the only installed backend, asks only if multiple non-native are present.
+  - `preferred_image_backend: codex-imagegen` — pin to Codex's built-in.
+  - `preferred_image_backend: baoyu-imagine` — pin to the baoyu-imagine skill.
+  - `preferred_image_backend: ask` — confirm backend every run.
+  - `preferred_type: infographic`, `preferred_style: notion`, `preferred_palette: macaron`, `language: zh`.
+  - `default_output_dir: imgs-subdir` — where to write generated images relative to the article.
@@ -61,8 +61,10 @@ Position defaults to bottom-right.
 header: "Style"
 question: "Default illustration style preference? Or type another style name or your custom style"
 options:
-  - label: "None (Recommended)"
-    description: "Auto-select based on content analysis"
+  - label: "sketch-notes (Recommended)"
+    description: "Warm cream paper, black hand-drawn lines, soft pastel blocks — educational infographic feel. Great default for most articles."
+  - label: "None"
+    description: "Auto-select based on content analysis (falls back to sketch-notes when no strong signal)"
  - label: "notion"
    description: "Minimalist hand-drawn line art"
  - label: "warm"
@@ -126,13 +128,13 @@ preferred_style:
  description: ""
 default_output_dir: imgs-subdir  # same-dir | imgs-subdir | illustrations-subdir | independent
 language: null
+preferred_image_backend: auto
 custom_styles: []
 ---
 ```

+`preferred_image_backend: auto` is the baked-in default — first-time setup does not ask about it. The `## Image Generation Tools` rule in SKILL.md then picks the runtime-native tool (Codex `imagegen`, Hermes `image_generate`, etc.) when available, and falls back to installed backends.
+
 ## Modifying Preferences Later

-Users can edit EXTEND.md directly or run setup again:
- Delete EXTEND.md to trigger setup
- Edit YAML frontmatter for quick changes
- Full schema: `config/preferences-schema.md`
+See the `## Changing Preferences` section in `SKILL.md` for the canonical list of common edits (pin backend, change defaults, retrigger setup). Full schema: `preferences-schema.md`.
@@ -26,6 +26,8 @@ language: null            # zh|en|ja|ko|auto

 default_output_dir: null  # same-dir|illustrations-subdir|independent

+preferred_image_backend: auto  # auto|ask|<backend-id>
+
 custom_styles:
  - name: my-style
    description: "Style description"
@@ -52,6 +54,7 @@ custom_styles:
 | `preferred_palette` | string | null | Palette override (macaron, warm, neon, or null) |
 | `language` | string | null | Output language (null = auto-detect) |
 | `default_output_dir` | enum | null | Output directory preference (null = ask each time) |
+| `preferred_image_backend` | string | `auto` | Image backend selection. `auto` = prefer runtime-native tool, fall back to the only installed backend, ask if multiple non-native are present. `ask` = always confirm on every run. `<backend-id>` (e.g., `codex-imagegen`, `baoyu-imagine`, `image_generate`) = pin this backend when available; fall back to `auto` when it isn't. Absent = `auto`. Resolution logic is documented in `SKILL.md`'s `## Image Generation Tools` section. |
 | `custom_styles` | array | [] | User-defined styles |

 ## Position Options
@@ -113,6 +116,8 @@ preferred_style:

 language: zh

+preferred_image_backend: codex-imagegen
+
 custom_styles:
  - name: corporate
    description: "Professional B2B style"
@@ -139,6 +139,39 @@ STYLE: [style characteristics]
 ASPECT: 16:9
 ```

+**Infographic + sketch-notes + macaron palette** (default / `hand-drawn-edu` preset):
+```
+Single-page hand-drawn educational infographic in a clean presentation style.
+Warm cream paper background, black hand-drawn lines with slight wobble, soft
+pastel color blocks. Feels simple, friendly, and easy to understand at a glance.
+Diagram-style visuals ONLY — no realistic or photographic images.
+
+PALETTE: macaron — soft pastel blocks on warm cream
+COLORS: Warm Cream background (#F5F0E8); Black (#1A1A1A) for ALL lines, text,
+        arrows, and doodles; section fills in Light Blue (#A8D8EA), Mint Green
+        (#B5E5CF), Lavender (#D5C6E0), Peach (#FFD5C2); Coral Red (#E8655A)
+        sparingly for one or two emphasis points only.
+
+LAYOUT (top → bottom):
+- TOP: Bold hand-lettered title, oversized, slightly wobbly, with an optional
+       decorative underline or small doodle.
+- MIDDLE: 2–6 rounded-rectangle info boxes arranged in a clean grid, row, or
+          radial pattern. Each box = one section, one pastel fill color, one
+          simple icon or sketchy cartoon element, one short keyword/phrase.
+          Hand-drawn arrows connect related zones.
+- BOTTOM: One short hand-lettered takeaway sentence summarizing the main idea.
+
+ELEMENTS: Rounded info boxes with clear sectioning, wavy/straight hand-drawn
+          arrows with small inline labels, simple icons and sketchy cartoon
+          elements (stick figures, tools, objects), small doodle decorations
+          (stars, sparkles, underlines, dots, asterisks) used sparingly.
+
+STYLE: Minimal, well-organized, airy. Color fills don't completely fill
+       outlines (slight "hand-painted" overshoot). ALL text hand-lettered —
+       no computer fonts. Short labels and keywords only, never long
+       paragraphs. Generous white space between sections.
+```
+
 **Infographic + vector-illustration**:
 ```
 Flat vector illustration infographic. Clean black outlines on all elements.
@@ -2,6 +2,10 @@

 `--preset X` expands to a type + style + optional palette combination. Users can override any dimension.

+## Default Preset
+
+When content analysis surfaces no strong signal (generic knowledge article, mixed-topic post, no clear data/comparison/narrative cue), recommend **`hand-drawn-edu`** as the primary option in Step 3 Q1. It is the warm, friendly educational-infographic default — safe for most articles and universally readable.
+
 ## By Category

 ### Technical & Engineering
@@ -23,7 +27,9 @@
 | `process-flow` | `flowchart` | `notion` | — | Workflow documentation, onboarding flows |
 | `warm-knowledge` | `infographic` | `vector-illustration` | `warm` | Product showcases, team intros, feature cards, brand content |
 | `edu-visual` | `infographic` | `vector-illustration` | `macaron` | Knowledge summaries, concept explainers, educational articles |
-| `hand-drawn-edu` | `flowchart` | `sketch-notes` | `macaron` | Hand-drawn educational diagrams, process explainers, onboarding visuals |
+| `hand-drawn-edu` | `infographic` | `sketch-notes` | `macaron` | **Default preset.** Hand-drawn educational infographic — warm cream paper, black lines, pastel blocks. Great for single-page explainers, concept summaries, onboarding, general knowledge articles |
+| `hand-drawn-edu-flow` | `flowchart` | `sketch-notes` | `macaron` | Hand-drawn process explainer — step-by-step workflow in the same warm educational style |
+| `hand-drawn-edu-compare` | `comparison` | `sketch-notes` | `macaron` | Hand-drawn side-by-side comparison in the warm educational style |
 | `ink-notes-compare` | `comparison` | `ink-notes` | `mono-ink` | Before/After essays, Traditional vs New, OS-style comparisons, mindset-shift narratives |
 | `ink-notes-flow` | `flowchart` | `ink-notes` | `mono-ink` | Professional process explainers, workforce pipelines, hand-drawn technical walkthroughs |
 | `ink-notes-framework` | `framework` | `ink-notes` | `mono-ink` | System analogies, command-center diagrams, architecture-as-metaphor, tech manifestos |
@@ -59,18 +65,19 @@ Use this table during Step 3 to recommend presets based on Step 2 content analys

 | Content Type (Step 2) | Primary Preset | Alternatives |
 |------------------------|----------------|--------------|
-| Technical | `tech-explainer` | `system-design`, `architecture` |
-| Tutorial | `tutorial` | `process-flow`, `knowledge-base`, `edu-visual` |
+| **General / No strong signal** | `hand-drawn-edu` | `edu-visual`, `knowledge-base` |
+| Education / Knowledge | `hand-drawn-edu` | `edu-visual`, `knowledge-base`, `tutorial` |
+| Tutorial | `hand-drawn-edu-flow` | `tutorial`, `process-flow`, `hand-drawn-edu` |
+| SaaS / Product | `hand-drawn-edu` | `saas-guide`, `knowledge-base`, `process-flow`, `warm-knowledge` |
+| Technical | `tech-explainer` | `system-design`, `architecture`, `hand-drawn-edu` |
 | Methodology / Framework | `system-design` | `architecture`, `process-flow` |
 | Data / Metrics | `data-report` | `versus`, `tech-explainer` |
-| Comparison / Review | `versus` | `business-compare`, `editorial-poster`, `ink-notes-compare` |
+| Comparison / Review | `versus` | `business-compare`, `hand-drawn-edu-compare`, `editorial-poster`, `ink-notes-compare` |
 | Manifesto / Mindset shift / Professional visual note | `ink-notes-compare` | `ink-notes-framework`, `ink-notes-flow` |
 | Narrative / Personal | `storytelling` | `lifestyle`, `evolution` |
 | Opinion / Editorial | `opinion-piece` | `cinematic`, `editorial-poster` |
 | Historical / Timeline | `history` | `evolution` |
 | Academic / Research | `science-paper` | `tech-explainer`, `data-report` |
-| SaaS / Product | `saas-guide` | `knowledge-base`, `process-flow`, `warm-knowledge` |
-| Education / Knowledge | `edu-visual` | `knowledge-base`, `tutorial`, `hand-drawn-edu` |

 ## Override Examples

@@ -6,15 +6,15 @@ Simplified style tier for quick selection:

 | Core Style | Maps To | Best For |
 |------------|---------|----------|
+| `hand-drawn` | sketch-notes | **Default.** Warm cream paper, black hand-drawn lines, pastel blocks — educational infographics, concept explainers, onboarding, general knowledge articles |
 | `vector` | vector-illustration | Knowledge articles, tutorials, tech content |
 | `minimal-flat` | notion | General, knowledge sharing, SaaS |
 | `sci-fi` | blueprint | AI, frontier tech, system design |
-| `hand-drawn` | sketch/warm | Relaxed, reflective, casual content |
 | `editorial` | editorial | Processes, data, journalism |
 | `scene` | warm/watercolor | Narratives, emotional, lifestyle |
 | `poster` | screen-print | Opinion, editorial, cultural, cinematic |

-Use Core Styles for most cases. See full Style Gallery below for granular control.
+Use Core Styles for most cases. **When no strong content signal is detected, default to `hand-drawn` (→ sketch-notes).** See full Style Gallery below for granular control.

 ---

@@ -50,42 +50,45 @@ Full specifications: `references/styles/<style>.md`

 ## Type × Style Compatibility Matrix

-| | vector-illustration | notion | warm | minimal | blueprint | watercolor | elegant | editorial | scientific | screen-print |
-|---|:---:|:---:|:---:|:---:|:---:|:---:|:---:|:---:|:---:|:---:|
-| infographic | ✓✓ | ✓✓ | ✓ | ✓✓ | ✓✓ | ✓ | ✓✓ | ✓✓ | ✓✓ | ✓ |
-| scene | ✓ | ✓ | ✓✓ | ✓ | ✗ | ✓✓ | ✓ | ✓ | ✗ | ✓✓ |
-| flowchart | ✓✓ | ✓✓ | ✓ | ✓ | ✓✓ | ✗ | ✓ | ✓✓ | ✓ | ✗ |
-| comparison | ✓✓ | ✓✓ | ✓ | ✓✓ | ✓ | ✓ | ✓✓ | ✓✓ | ✓ | ✓ |
-| framework | ✓✓ | ✓✓ | ✓ | ✓✓ | ✓✓ | ✗ | ✓✓ | ✓ | ✓✓ | ✓ |
-| timeline | ✓ | ✓✓ | ✓ | ✓ | ✓ | ✓✓ | ✓✓ | ✓✓ | ✓ | ✓ |
+| | sketch-notes | vector-illustration | notion | warm | minimal | blueprint | watercolor | elegant | editorial | scientific | screen-print |
+|---|:---:|:---:|:---:|:---:|:---:|:---:|:---:|:---:|:---:|:---:|:---:|
+| infographic | ✓✓ | ✓✓ | ✓✓ | ✓ | ✓✓ | ✓✓ | ✓ | ✓✓ | ✓✓ | ✓✓ | ✓ |
+| scene | ✗ | ✓ | ✓ | ✓✓ | ✓ | ✗ | ✓✓ | ✓ | ✓ | ✗ | ✓✓ |
+| flowchart | ✓✓ | ✓✓ | ✓✓ | ✓ | ✓ | ✓✓ | ✗ | ✓ | ✓✓ | ✓ | ✗ |
+| comparison | ✓✓ | ✓✓ | ✓✓ | ✓ | ✓✓ | ✓ | ✓ | ✓✓ | ✓✓ | ✓ | ✓ |
+| framework | ✓✓ | ✓✓ | ✓✓ | ✓ | ✓✓ | ✓✓ | ✗ | ✓✓ | ✓ | ✓✓ | ✓ |
+| timeline | ✓ | ✓ | ✓✓ | ✓ | ✓ | ✓ | ✓✓ | ✓✓ | ✓✓ | ✓ | ✓ |

 ✓✓ = highly recommended | ✓ = compatible | ✗ = not recommended

 ## Auto Selection by Type

+When no content signal matches strongly, `sketch-notes` is the default primary for every diagrammatic type. Only override with another primary when the content analysis in Step 2 surfaces a clear signal (technical/data/narrative/opinion).
+
 | Type | Primary Style | Secondary Styles |
 |------|---------------|------------------|
-| infographic | vector-illustration | notion, blueprint, editorial |
+| infographic | sketch-notes | vector-illustration, notion, blueprint, editorial |
 | scene | warm | watercolor, elegant |
-| flowchart | vector-illustration | notion, blueprint |
-| comparison | vector-illustration | notion, elegant |
-| framework | blueprint | vector-illustration, notion |
-| timeline | elegant | warm, editorial |
+| flowchart | sketch-notes | vector-illustration, notion, blueprint |
+| comparison | sketch-notes | vector-illustration, notion, elegant |
+| framework | sketch-notes | blueprint, vector-illustration, notion |
+| timeline | elegant | sketch-notes, warm, editorial |

 ## Auto Selection by Content Signals

 | Content Signals | Recommended Type | Recommended Style |
 |-----------------|------------------|-------------------|
+| **(no strong signal / general article)** | **infographic** | **sketch-notes** |
+| Knowledge, concept, tutorial, learning, guide, onboarding | infographic | sketch-notes, vector-illustration, notion |
+| Productivity, SaaS, tool, app, software | infographic | sketch-notes, notion, vector-illustration |
+| How-to, steps, workflow, process, tutorial | flowchart | sketch-notes, vector-illustration, notion |
 | API, metrics, data, comparison, numbers | infographic | blueprint, vector-illustration |
-| Knowledge, concept, tutorial, learning, guide | infographic | vector-illustration, notion |
-| Tech, AI, programming, development, code | infographic | vector-illustration, blueprint |
-| How-to, steps, workflow, process, tutorial | flowchart | vector-illustration, notion |
-| Framework, model, architecture, principles | framework | blueprint, vector-illustration |
-| vs, pros/cons, before/after, alternatives | comparison | vector-illustration, notion |
+| Tech, AI, programming, development, code | infographic | vector-illustration, blueprint, sketch-notes |
+| Framework, model, architecture, principles | framework | blueprint, vector-illustration, sketch-notes |
+| vs, pros/cons, before/after, alternatives | comparison | vector-illustration, notion, sketch-notes |
 | Manifesto, mindset shift, workforce, OS, whiteboard, professional visual note | comparison / framework | ink-notes |
 | Story, emotion, journey, experience, personal | scene | warm, watercolor |
 | History, timeline, progress, evolution | timeline | elegant, warm |
-| Productivity, SaaS, tool, app, software | infographic | notion, vector-illustration |
 | Business, professional, strategy, corporate | framework | elegant |
 | Opinion, editorial, culture, philosophy, cinematic, dramatic, poster | scene | screen-print |
 | Biology, chemistry, medical, scientific | infographic | scientific |
@@ -93,6 +96,15 @@ Full specifications: `references/styles/<style>.md`

 ## Style Characteristics by Type

+### infographic + sketch-notes (default)
+- Warm cream paper background, black hand-drawn lines with slight wobble
+- 2–6 rounded pastel info boxes (light blue / mint / lavender / peach)
+- Bold hand-lettered title at the top
+- Short keyword labels, simple icons, small doodles (stars, underlines, sparkles)
+- One-line hand-lettered takeaway sentence at the bottom
+- Airy, minimal, diagram-style — never realistic
+- Perfect for single-page educational explainers and concept summaries
+
 ### infographic + vector-illustration
 - Clean flat vector shapes, bold geometric forms
 - Vibrant but harmonious color palette
@@ -1,56 +1,91 @@
 # sketch-notes

-Soft hand-drawn illustration style with warm, educational feel
+Hand-drawn educational infographic style with warm cream paper, black hand-drawn lines, and soft pastel section blocks. Optimized for single-page visual explainers.

 ## Design Aesthetic

-Hand-drawn feel with soft, relaxed brush strokes. Fresh, refined style with minimalist editorial approach. Emphasis on precision, clarity and intelligent elegance while prioritizing warmth, approachability and friendliness.
+Hand-drawn educational infographic in a clean presentation style. Feels like a visual explainer slide: simple, friendly, and easy to understand at a glance. Bold handwritten-style title at the top, clearly sectioned content in the middle with rounded boxes and small doodles, and one short takeaway sentence at the bottom. Neat, airy, and visually similar to a hand-drawn concept diagram — never realistic or photographic.

 ## Background

- Color: Warm Off-White (#FAF8F0)
- Texture: Subtle paper grain, warm tone
+- Color: Warm Cream Paper (#F5F0E8) — preferred; fallback Warm Off-White (#FAF8F0)
+- Texture: Subtle warm paper grain, matte finish, no gloss

 ## Color Palette

+Default sketch-notes palette is the **macaron** pastel set. Lines are always black; pastel blocks are used only as rounded card fills for information sections.
+
 | Role | Color | Hex | Usage |
 |------|-------|-----|-------|
-| Background | Warm Off-White | #FAF8F0 | Primary background |
-| Primary Text | Deep Charcoal | #2C3E50 | Main elements |
-| Alt Text | Deep Brown | #4A4A4A | Secondary elements |
-| Accent 1 | Soft Orange | #F4A261 | Highlights, emphasis |
-| Accent 2 | Mustard Yellow | #E9C46A | Secondary highlights |
-| Accent 3 | Sage Green | #87A96B | Nature, growth concepts |
-| Accent 4 | Light Blue | #7EC8E3 | Tech, digital elements |
-| Accent 5 | Red Brown | #A0522D | Earthy elements |
+| Background | Warm Cream | #F5F0E8 | Paper background |
+| Primary Ink | Black | #1A1A1A | ALL outlines, text, arrows, doodles |
+| Block Blue | Light Blue | #A8D8EA | Info block fill (cool / tech) |
+| Block Mint | Mint Green | #B5E5CF | Info block fill (growth / positive) |
+| Block Lavender | Lavender | #D5C6E0 | Info block fill (concept / abstract) |
+| Block Peach | Peach | #FFD5C2 | Info block fill (warm / human) |
+| Accent | Coral Red | #E8655A | One or two emphasis points only |
+| Muted Text | Warm Gray | #6B6B6B | Small annotations |
+
+Use **4 pastel block colors max** per image, one color per section. Black ink does all the structural line work.

 ## Visual Elements

- Connection lines with hand-drawn wavy feel
- Conceptual abstract icons illustrating ideas
- Color fills don't completely fill outlines (hand-painted feel)
- Simple geometric shapes with rounded corners
- Arrows and pointers with sketchy style
- Doodle decorations: stars, spirals, underlines
+- Bold hand-lettered title at the top (oversized, slightly wobbly)
+- Rounded-rectangle info boxes with clear sectioning (2–6 zones)
+- Short keyword labels inside boxes — never long paragraphs
+- Simple icons and sketchy cartoon elements (stick figures, tools, objects) to explain each idea
+- Hand-drawn arrows (straight, curved, or wavy) connecting related zones
+- Small doodle decorations: stars, sparkles, underlines, dots, asterisks — used sparingly for emphasis
+- Single-line hand-lettered takeaway sentence at the bottom
+- Color fills do not completely fill outlines (slight "hand-painted" overshoot/undershoot)
+- Generous white space between sections — airy, never crowded
+
+## Layout Guidelines
+
+Canonical single-page layout (16:9 or 4:3):
+
+1. **Top (10–15%)** — Bold hand-lettered title, optionally with a small decorative underline or doodle.
+2. **Middle (70–80%)** — 2–6 rounded pastel info boxes arranged in a clear grid, row, or radial pattern. Each box = one section, one color, one icon, one keyword/phrase.
+3. **Bottom (10–15%)** — One short hand-lettered takeaway sentence summarizing the core insight.
+
+Keep margins generous. Aim for breathing room around every element.

 ## Style Rules

 ### Do

- Keep layouts open and well-structured
- Emphasize information hierarchy
- Use hand-drawn quality for all elements
- Allow imperfection (slight wobbles add character)
- Layer elements with subtle overlaps
+- Use warm cream paper background (no pure white)
+- Use black hand-drawn lines for ALL structural elements
+- Use soft pastel blocks (blue / mint / lavender / peach) for section fills
+- Keep text to short keywords and phrases only
+- Include a bold handwritten title at the top
+- Include a short takeaway sentence at the bottom
+- Use diagram-style visuals (icons, doodles, simple shapes)
+- Allow slight wobble — hand-drawn imperfection is the point
+- Maintain clear sectioning with rounded boxes

 ### Don't

- Use perfect geometric shapes
- Create photorealistic elements
- Overcrowd with too many elements
- Use pure white backgrounds
- Make it look computer-generated
+- Use pure white backgrounds (that's `ink-notes`' territory)
+- Render realistic or photographic images — this style is diagram-only
+- Fill zones with gradients, shadows, or digital effects
+- Use long paragraphs of text — keywords only
+- Use computer-generated / sans-serif body fonts — ALL text must be hand-lettered
+- Use more than 4 pastel block colors per image
+- Overcrowd the canvas — keep it airy and minimal
+- Use perfect geometric shapes — preserve the hand-drawn wobble
+
+## Type Compatibility
+
+| Type | Rating | Notes |
+|------|--------|-------|
+| infographic | ✓✓ | **Best fit** — single-page visual explainers, concept summaries, educational slides |
+| framework | ✓✓ | Labeled zones and connectors render well |
+| flowchart | ✓✓ | Rounded step boxes with wavy arrows |
+| comparison | ✓✓ | Two pastel blocks side by side; prefer `ink-notes` for strict Before/After contrasts |
+| timeline | ✓ | Hand-drawn horizontal arrow with milestone cards |
+| scene | ✗ | Not recommended — too diagrammatic |

 ## Best For

-Educational content, knowledge sharing, technical explanations, tutorials, onboarding materials, friendly articles
+Educational content, knowledge sharing, concept explainers, tutorials, onboarding materials, product walkthroughs, single-page visual summaries, "how things work" posts, friendly technical articles
@@ -176,6 +176,8 @@ Based on Step 2 content analysis, recommend a preset first (sets both type & sty
 - [Alternative preset] — [brief]
 - Or choose type manually: infographic / scene / flowchart / comparison / framework / timeline / mixed

+**Default**: if Step 2 found no strong content signal, the recommended preset MUST be `hand-drawn-edu` (infographic + sketch-notes + macaron — warm cream paper, black hand-drawn lines, soft pastel blocks). This is the universal fallback.
+
 **If user picks a preset → skip Q3** (type & style both resolved).
 **If user picks a type → Q3 is REQUIRED.**

@@ -203,13 +205,15 @@ If no `preferred_style` (present Core Styles first):

 | Core Style | Maps To | Best For |
 |------------|---------|----------|
+| `hand-drawn` | sketch-notes | **Default.** Warm cream paper, black hand-drawn lines, pastel blocks — educational infographics, concept explainers, onboarding, general knowledge articles |
 | `minimal-flat` | notion | General, knowledge sharing, SaaS |
 | `sci-fi` | blueprint | AI, frontier tech, system design |
-| `hand-drawn` | sketch/warm | Relaxed, reflective, casual |
 | `editorial` | editorial | Processes, data, journalism |
 | `scene` | warm/watercolor | Narratives, emotional, lifestyle |
 | `poster` | screen-print | Opinion, editorial, cultural, cinematic |

+**Default recommendation**: when Step 2 surfaces no strong content signal, recommend **`hand-drawn-edu`** preset (→ infographic + sketch-notes + macaron) as the primary option in Q1. When the user picks a type manually without a preferred_style, recommend `sketch-notes` first in Q3.
+
 Style selection based on Type × Style compatibility matrix (styles.md).
 **In Step 5.1**, read `styles/<style>.md` for visual elements and rendering rules.

@@ -27,11 +27,17 @@ Concrete `AskUserQuestion` references below are examples — substitute the loca

 ## Image Generation Tools

-When this skill needs to render an image:
+When this skill needs to render an image, resolve the backend in this order:

- **Use whatever image-generation tool or skill is available** in the current runtime — e.g., Codex `imagegen`, Hermes `image_generate`, `baoyu-imagine`, or any equivalent the user has installed.
- **If multiple are available**, ask the user **once** at the start which to use (batch with any other initial questions).
- **If none are available**, tell the user and ask how to proceed.
+1. **Current-request override** — if the user names a specific backend in the current message, use it.
+2. **Saved preference** — if `EXTEND.md` sets `preferred_image_backend` to a backend available right now, use it.
+3. **Auto-select** (when the preference is `auto`, unset, or the pinned backend isn't available):
+   - If the current runtime exposes a native image tool (e.g., Codex `imagegen`, Hermes `image_generate`), use it. Runtime-native tools are preferred by default — agents that know their own tool inventory should surface the native one here.
+   - Otherwise, if exactly one non-native backend is installed (e.g., `baoyu-imagine`), use it.
+   - Otherwise (multiple non-native backends with no runtime-native tool), ask the user once — batch with any other initial questions.
+4. **If none are available**, tell the user and ask how to proceed.
+
+Setting `preferred_image_backend: ask` forces the step-3 prompt every run regardless of available backends. Users change the pinned backend via the `## Changing Preferences` section below.

 **Prompt file requirement (hard)**: write each image's full, final prompt to a standalone file under `prompts/` (naming: `NN-{type}-[slug].md`) BEFORE invoking any backend. The backend receives the prompt file (or its content); the file is the reproducibility record and lets you switch backends without regenerating prompts.

@@ -295,3 +301,16 @@ If EXTEND.md is not found, first-time setup is **blocking** — complete it befo
 - **Step 7.1 character sheet** - recommended for multi-page comics, optional for simple presets
 - **Step 7.2 character reference** - use `--ref` if sheet exists; compress/convert on failure; fall back to prompt-only
 - Watermark/language configured once in EXTEND.md
+
+## Changing Preferences
+
+EXTEND.md lives at `.baoyu-skills/baoyu-comic/EXTEND.md` (project) or `~/.baoyu-skills/baoyu-comic/EXTEND.md` (user). Three ways to change it:
+
+- **Edit directly** — open EXTEND.md and change fields. Full schema: `references/config/preferences-schema.md`.
+- **Reconfigure interactively** — delete EXTEND.md (or ask "reconfigure baoyu-comic preferences" / "重新配置"). The next run re-triggers first-time setup.
+- **Common one-line edits**:
+  - `preferred_image_backend: auto` — default; runtime-native tool wins, falls back to the only installed backend, asks only if multiple non-native are present.
+  - `preferred_image_backend: codex-imagegen` — pin to Codex's built-in.
+  - `preferred_image_backend: baoyu-imagine` — pin to the baoyu-imagine skill.
+  - `preferred_image_backend: ask` — confirm backend every run.
+  - `watermark.enabled: true`, `preferred_art`, `preferred_tone`, `preferred_layout`, `language` — shift the auto-selection defaults and cosmetic choices.
@@ -142,13 +142,13 @@ preferred_tone: [selected tone or null]
 preferred_layout: null
 preferred_aspect: null
 language: [selected or null]
+preferred_image_backend: auto
 character_presets: []
 ---
 ```

+`preferred_image_backend: auto` is the baked-in default — first-time setup does not ask about it. The `## Image Generation Tools` rule in SKILL.md then picks the runtime-native tool (Codex `imagegen`, Hermes `image_generate`, etc.) when available, and falls back to installed backends.
+
 ## Modifying Preferences Later

-Users can edit EXTEND.md directly or run setup again:
- Delete EXTEND.md to trigger setup
- Edit YAML frontmatter for quick changes
- Full schema: `config/preferences-schema.md`
+See the `## Changing Preferences` section in `SKILL.md` for the canonical list of common edits (pin backend, change defaults, retrigger setup). Full schema: `config/preferences-schema.md`.
@@ -23,6 +23,8 @@ preferred_aspect: null    # 3:4|4:3|16:9

 language: null            # zh|en|ja|ko|auto

+preferred_image_backend: auto  # auto|ask|<backend-id>
+
 character_presets:
  - name: my-characters
    roles:
@@ -46,6 +48,7 @@ character_presets:
 | `preferred_layout` | string | null | Layout preference or null |
 | `preferred_aspect` | string | null | Aspect ratio (3:4, 4:3, 16:9) |
 | `language` | string | null | Output language (null = auto-detect) |
+| `preferred_image_backend` | string | `auto` | Image backend selection. `auto` = prefer runtime-native tool, fall back to the only installed backend, ask if multiple non-native are present. `ask` = always confirm on every run. `<backend-id>` (e.g., `codex-imagegen`, `baoyu-imagine`, `image_generate`) = pin this backend when available; fall back to `auto` when it isn't. Absent = `auto`. Resolution logic is documented in `SKILL.md`'s `## Image Generation Tools` section. |
 | `character_presets` | array | [] | Preset character roles for styles like ohmsha |

 ## Art Style Options
@@ -122,6 +125,8 @@ preferred_aspect: "3:4"

 language: zh

+preferred_image_backend: codex-imagegen
+
 character_presets:
  - name: tech-tutorial
    roles:
@@ -1,7 +1,7 @@
 ---
 name: baoyu-cover-image
 description: Generates article cover images with 5 dimensions (type, palette, rendering, text, mood) combining 11 color palettes and 7 rendering styles. Supports cinematic (2.35:1), widescreen (16:9), and square (1:1) aspects. Use when user asks to "generate cover image", "create article cover", or "make cover".
-version: 1.56.1
+version: 1.56.2
 metadata:
  openclaw:
    homepage: https://github.com/JimLiu/baoyu-skills#baoyu-cover-image
@@ -23,16 +23,31 @@ Concrete `AskUserQuestion` references below are examples — substitute the loca

 ## Image Generation Tools

-When this skill needs to render an image:
+When this skill needs to render an image, resolve the backend in this order:

- **Use whatever image-generation tool or skill is available** in the current runtime — e.g., Codex `imagegen`, Hermes `image_generate`, `baoyu-imagine`, or any equivalent the user has installed.
- **If multiple are available**, ask the user **once** at the start which to use (batch with any other initial questions).
- **If none are available**, tell the user and ask how to proceed.
+1. **Current-request override** — if the user names a specific backend in the current message, use it.
+2. **Saved preference** — if `EXTEND.md` sets `preferred_image_backend` to a backend available right now, use it.
+3. **Auto-select** (when the preference is `auto`, unset, or the pinned backend isn't available):
+   - If the current runtime exposes a native image tool (e.g., Codex `imagegen`, Hermes `image_generate`), use it. Runtime-native tools are preferred by default — agents that know their own tool inventory should surface the native one here.
+   - Otherwise, if exactly one non-native backend is installed (e.g., `baoyu-imagine`), use it.
+   - Otherwise (multiple non-native backends with no runtime-native tool), ask the user once — batch with any other initial questions.
+4. **If none are available**, tell the user and ask how to proceed.
+
+Setting `preferred_image_backend: ask` forces the step-3 prompt every run regardless of available backends. Users change the pinned backend via the `## Changing Preferences` section below.

 **Prompt file requirement (hard)**: write each image's full, final prompt to a standalone file under `prompts/` (naming: `NN-{type}-[slug].md`) BEFORE invoking any backend. The backend receives the prompt file (or its content); the file is the reproducibility record and lets you switch backends without regenerating prompts.

 Concrete tool names (`imagegen`, `image_generate`, `baoyu-imagine`) above are examples — substitute the local equivalents under the same rule.

+## Confirmation Policy
+
+Default behavior: **confirm before generation**.
+
+- Treat explicit skill invocation, a file path, matched keywords/presets, `EXTEND.md` defaults, and any documented auto-selection as **recommendation inputs only**. None of them authorizes skipping confirmation.
+- Do **not** start Step 3 or Step 4 until the user confirms the dimensions / aspect / language / backend choices.
+- Skip confirmation only when the current request explicitly says to do so, for example: `--quick`, "直接生成", "不用确认", "跳过确认", "按默认出图", or equivalent wording. `quick_mode: true` in `EXTEND.md` counts as a standing explicit opt-out — set it only when you want every run to skip Step 2.
+- If confirmation is skipped explicitly, state the assumed dimensions / aspect / language / backend in the next user-facing update before generating.
+
 ## Options

 | Option | Description |
@@ -164,6 +179,8 @@ See [reference-images.md](references/workflow/reference-images.md) for full deci

 ### Step 2: Confirm Options ⚠️

+**Hard gate**: this step is mandatory per the [Confirmation Policy](#confirmation-policy) — Steps 3–4 cannot start until the user confirms here (or explicitly opts out with `--quick` / `quick_mode: true` / equivalent wording in the current request).
+
 **MUST use `AskUserQuestion` tool** to present options as interactive selection — NOT plain text tables. Present up to 4 questions in a single `AskUserQuestion` call (Type, Palette, Rendering, Font + Settings). Each question shows the recommended option first with reason, followed by alternatives.

 Full confirmation flow and question format: [references/workflow/confirm-options.md](references/workflow/confirm-options.md)
@@ -228,13 +245,18 @@ Files:
 - **Characters**: Simplified silhouettes; NO realistic humans
 - **Title**: Use exact title from user/source; never invent

-## Extension Support
+## Changing Preferences

-Custom configurations via EXTEND.md. See **Step 0** for paths.
+EXTEND.md lives at the path noted in **Step 0**. Three ways to change it:

-Supports: Watermark | Preferred dimensions | Default aspect/output | Quick mode | Custom palettes | Language
-
-Schema: [references/config/preferences-schema.md](references/config/preferences-schema.md)
+- **Edit directly** — open EXTEND.md and change fields. Full schema: [references/config/preferences-schema.md](references/config/preferences-schema.md).
+- **Reconfigure interactively** — delete EXTEND.md (or ask "reconfigure baoyu-cover-image preferences" / "重新配置"). The next run re-triggers first-time setup.
+- **Common one-line edits**:
+  - `preferred_image_backend: auto` — default; runtime-native tool wins, falls back to the only installed backend, asks only if multiple non-native are present.
+  - `preferred_image_backend: codex-imagegen` — pin to Codex's built-in.
+  - `preferred_image_backend: baoyu-imagine` — pin to the baoyu-imagine skill.
+  - `preferred_image_backend: ask` — confirm backend every run.
+  - `watermark.enabled: true`, `preferred_type`, `preferred_palette`, `preferred_rendering`, `default_aspect`, `quick_mode: true`, `language` — shift the auto-selection defaults and confirmation flow.

 ## References

@@ -188,15 +188,15 @@ default_aspect: [16:9/2.35:1/1:1/3:4]
 default_output_dir: [independent/same-dir/imgs-subdir]
 quick_mode: [true/false]
 language: null
+preferred_image_backend: auto
 custom_palettes: []
 ---
 ```

+`preferred_image_backend: auto` is the baked-in default — first-time setup does not ask about it. The `## Image Generation Tools` rule in SKILL.md then picks the runtime-native tool (Codex `imagegen`, Hermes `image_generate`, etc.) when available, and falls back to installed backends.
+
 ## Modifying Preferences Later

-Users can edit EXTEND.md directly or run setup again:
- Delete EXTEND.md to trigger setup
- Edit YAML frontmatter for quick changes
- Full schema: `preferences-schema.md`
+See the `## Changing Preferences` section in `SKILL.md` for the canonical list of common edits (pin backend, change defaults, retrigger setup). Full schema: `preferences-schema.md`.

-**EXTEND.md Supports**: Watermark | Preferred type | Preferred palette | Preferred rendering | Preferred text | Preferred mood | Default aspect ratio | Default output directory | Quick mode | Custom palette definitions | Language preference
+**EXTEND.md Supports**: Watermark | Preferred type | Preferred palette | Preferred rendering | Preferred text | Preferred mood | Default aspect ratio | Default output directory | Quick mode | Image backend preference | Custom palette definitions | Language preference
@@ -32,6 +32,8 @@ quick_mode: false         # Skip confirmation when true

 language: null            # zh|en|ja|ko|auto (null = auto-detect)

+preferred_image_backend: auto  # auto|ask|<backend-id>
+
 custom_palettes:
  - name: my-palette
    description: "Palette description"
@@ -60,6 +62,7 @@ custom_palettes:
 | `default_aspect` | string | "2.35:1" | Default aspect ratio |
 | `quick_mode` | bool | false | Skip confirmation step |
 | `language` | string | null | Output language (null = auto-detect) |
+| `preferred_image_backend` | string | `auto` | Image backend selection. `auto` = prefer runtime-native tool, fall back to the only installed backend, ask if multiple non-native are present. `ask` = always confirm on every run. `<backend-id>` (e.g., `codex-imagegen`, `baoyu-imagine`, `image_generate`) = pin this backend when available; fall back to `auto` when it isn't. Absent = `auto`. Resolution logic is documented in `SKILL.md`'s `## Image Generation Tools` section. |
 | `custom_palettes` | array | [] | User-defined palettes |

 ## Type Options
@@ -187,6 +190,8 @@ quick_mode: true

 language: en

+preferred_image_backend: codex-imagegen
+
 custom_palettes:
  - name: corporate-tech
    description: "Professional B2B tech palette"
@@ -1,7 +1,7 @@
 ---
 name: baoyu-image-cards
 description: Generates infographic image card series with 12 visual styles, 8 layouts, and 3 color palettes. Breaks content into 1-10 cartoon-style image cards optimized for social media engagement. Use when user mentions "小红书图片", "小红书种草", "小绿书", "微信图文", "微信贴图", "image cards", "图片卡片", or wants social media infographic series.
-version: 1.56.1
+version: 1.56.2
 metadata:
  openclaw:
    homepage: https://github.com/JimLiu/baoyu-skills#baoyu-image-cards
@@ -23,14 +23,31 @@ Concrete `AskUserQuestion` references below are examples — substitute the loca

 ## Image Generation Tools

-When this skill needs to render an image:
+When this skill needs to render an image, resolve the backend in this order:

- **Use whatever image-generation tool or skill is available** in the current runtime — e.g., Codex `imagegen`, Hermes `image_generate`, `baoyu-imagine`, or any equivalent the user has installed.
- **If multiple are available**, ask the user **once** at the start which to use (batch with any other initial questions).
- **If none are available**, tell the user and ask how to proceed.
+1. **Current-request override** — if the user names a specific backend in the current message, use it.
+2. **Saved preference** — if `EXTEND.md` sets `preferred_image_backend` to a backend available right now, use it.
+3. **Auto-select** (when the preference is `auto`, unset, or the pinned backend isn't available):
+   - If the current runtime exposes a native image tool (e.g., Codex `imagegen`, Hermes `image_generate`), use it. Runtime-native tools are preferred by default — agents that know their own tool inventory should surface the native one here.
+   - Otherwise, if exactly one non-native backend is installed (e.g., `baoyu-imagine`), use it.
+   - Otherwise (multiple non-native backends with no runtime-native tool), ask the user once — batch with any other initial questions.
+4. **If none are available**, tell the user and ask how to proceed.
+
+Setting `preferred_image_backend: ask` forces the step-3 prompt every run regardless of available backends. Users change the pinned backend via the `## Changing Preferences` section below.

 **Prompt file requirement (hard)**: write each image's full, final prompt to a standalone file under `prompts/` (naming: `NN-{type}-[slug].md`) BEFORE invoking any backend. The file is the reproducibility record and lets you switch backends without regenerating prompts.

+Concrete tool names (`imagegen`, `image_generate`, `baoyu-imagine`) above are examples — substitute the local equivalents under the same rule.
+
+## Confirmation Policy
+
+Default behavior: **confirm before generation**.
+
+- Treat explicit skill invocation, a file path, matched signals/presets, and `EXTEND.md` defaults as **recommendation inputs only**. None of them authorizes skipping confirmation.
+- Do **not** start Step 3 until the user completes Step 2.
+- Skip confirmation only when the current request explicitly says to do so, for example: `--yes`, "直接生成", "不用确认", "跳过确认", "按默认出图", or equivalent wording.
+- If confirmation is skipped explicitly, state the assumed strategy / style / layout / palette / count / backend in the next user-facing update before generating.
+
 ## Language

 Respond in the user's language across questions, progress, errors, and completion summary. Keep technical tokens (style names, file paths, code) in English.
@@ -291,6 +308,8 @@ Check these paths in order; first hit wins:

 ### Step 2: Smart Confirm ⚠️ REQUIRED

+**Hard gate**: this step is mandatory per the [Confirmation Policy](#confirmation-policy) — Step 3 cannot start until the user confirms here (or explicitly opts out with `--yes` / equivalent wording in the current request).
+
 Goal: present the auto-recommended plan and let the user confirm or adjust. Skip this step entirely under `--yes` — proceed with Path A using the analysis and any CLI overrides.

 **Display summary** before asking:
@@ -417,4 +436,16 @@ Always update the prompt file before regenerating — it's the source of truth a
 - For sensitive public figures, use stylized cartoon alternatives.
 - Smart Confirm (Step 2) is required; Detailed mode adds a second confirmation (2a + 2c).

-Custom configurations via EXTEND.md. See Step 0 for paths and schema.
+## Changing Preferences
+
+EXTEND.md lives at the first matching path listed in Step 0. Three ways to change it:
+
+- **Edit directly** — open EXTEND.md and change fields. Full schema: `references/config/preferences-schema.md`.
+- **Reconfigure interactively** — delete EXTEND.md (or ask "reconfigure baoyu-image-cards preferences" / "重新配置"). The next run re-triggers first-time setup.
+- **Common one-line edits**:
+  - `preferred_image_backend: auto` — default; runtime-native tool wins, falls back to the only installed backend, asks only if multiple non-native are present.
+  - `preferred_image_backend: codex-imagegen` — pin to Codex's built-in.
+  - `preferred_image_backend: baoyu-imagine` — pin to the baoyu-imagine skill.
+  - `preferred_image_backend: ask` — confirm backend every run.
+  - `preferred_style: notion`, `preferred_layout: dense`, `preferred_palette: macaron`, `language: zh`.
+  - `watermark.enabled: true` + `watermark.content: "@handle"` — add a watermark.
@@ -110,13 +110,13 @@ preferred_style:
  description: ""
 preferred_layout: null
 language: null
+preferred_image_backend: auto
 custom_styles: []
 ---
 ```

+`preferred_image_backend: auto` is the baked-in default — first-time setup does not ask about it. The `## Image Generation Tools` rule in SKILL.md then picks the runtime-native tool (Codex `imagegen`, Hermes `image_generate`, etc.) when available, and falls back to installed backends.
+
 ## Modifying Preferences Later

-Users can edit EXTEND.md directly or run setup again:
- Delete EXTEND.md to trigger setup
- Edit YAML frontmatter for quick changes
- Full schema: `config/preferences-schema.md`
+See the `## Changing Preferences` section in `SKILL.md` for the canonical list of common edits (pin backend, change defaults, retrigger setup). Full schema: `preferences-schema.md`.
@@ -24,6 +24,8 @@ preferred_layout: null    # sparse|balanced|dense|list|comparison|flow

 language: null            # zh|en|ja|ko|auto

+preferred_image_backend: auto  # auto|ask|<backend-id>
+
 custom_styles:
  - name: my-style
    description: "Style description"
@@ -49,6 +51,7 @@ custom_styles:
 | `preferred_style.description` | string | "" | Custom notes/override |
 | `preferred_layout` | string | null | Layout preference or null |
 | `language` | string | null | Output language (null = auto-detect) |
+| `preferred_image_backend` | string | `auto` | Image backend selection. `auto` = prefer runtime-native tool, fall back to the only installed backend, ask if multiple non-native are present. `ask` = always confirm on every run. `<backend-id>` (e.g., `codex-imagegen`, `baoyu-imagine`, `image_generate`) = pin this backend when available; fall back to `auto` when it isn't. Absent = `auto`. Resolution logic is documented in `SKILL.md`'s `## Image Generation Tools` section. |
 | `custom_styles` | array | [] | User-defined styles |

 ## Position Options
@@ -104,6 +107,8 @@ preferred_layout: dense

 language: zh

+preferred_image_backend: codex-imagegen
+
 custom_styles:
  - name: corporate
    description: "Professional B2B style"
@@ -1,7 +1,7 @@
 ---
 name: baoyu-imagine
-description: AI image generation with OpenAI, Azure OpenAI, Google, OpenRouter, DashScope, Z.AI GLM-Image, MiniMax, Jimeng, Seedream and Replicate APIs. Supports text-to-image, reference images, aspect ratios, and batch generation from saved prompt files. Sequential by default; use batch parallel generation when the user already has multiple prompts or wants stable multi-image throughput. Use when user asks to generate, create, or draw images.
-version: 1.57.0
+description: AI image generation with OpenAI GPT Image 2, Azure OpenAI, Google, OpenRouter, DashScope, Z.AI GLM-Image, MiniMax, Jimeng, Seedream and Replicate APIs. Supports text-to-image, reference images, aspect ratios, and batch generation from saved prompt files. Sequential by default; use batch parallel generation when the user already has multiple prompts or wants stable multi-image throughput. Use when user asks to generate, create, or draw images.
+version: 1.58.0
 metadata:
  openclaw:
    homepage: https://github.com/JimLiu/baoyu-skills#baoyu-imagine
@@ -13,7 +13,7 @@ metadata:

 # Image Generation (AI SDK)

-Official API-based image generation. Supports OpenAI, Azure OpenAI, Google, OpenRouter, DashScope (阿里通义万象), Z.AI GLM-Image, MiniMax, Jimeng (即梦), Seedream (豆包) and Replicate.
+Official API-based image generation. Supports OpenAI GPT Image 2, Azure OpenAI, Google, OpenRouter, DashScope (阿里通义万象), Z.AI GLM-Image, MiniMax, Jimeng (即梦), Seedream (豆包) and Replicate.

 ## User Input Tools

@@ -68,6 +68,9 @@ ${BUN_X} {baseDir}/scripts/main.ts --prompt "Make blue" --image out.png --ref so
 # Specific provider
 ${BUN_X} {baseDir}/scripts/main.ts --prompt "A cat" --image out.png --provider dashscope --model qwen-image-2.0-pro

+# OpenAI GPT Image 2
+${BUN_X} {baseDir}/scripts/main.ts --prompt "A cat" --image out.png --provider openai --model gpt-image-2
+
 # Batch mode
 ${BUN_X} {baseDir}/scripts/main.ts --batchfile batch.json --jobs 4
 ```
@@ -84,11 +87,11 @@ ${BUN_X} {baseDir}/scripts/main.ts --batchfile batch.json --jobs 4
 | `--provider google\|openai\|azure\|openrouter\|dashscope\|zai\|minimax\|jimeng\|seedream\|replicate` | Force provider (default: auto-detect) |
 | `--model <id>`, `-m` | Model ID — see provider references for defaults and allowed values |
 | `--ar <ratio>` | Aspect ratio (`16:9`, `1:1`, `4:3`, …) |
-| `--size <WxH>` | Explicit size (e.g., `1024x1024`) |
+| `--size <WxH>` | Explicit size (e.g., `1024x1024`; for `gpt-image-2`, width/height must be multiples of 16, max edge 3840px, ratio no wider than 3:1) |
 | `--quality normal\|2k` | Quality preset (default: `2k`) |
 | `--imageSize 1K\|2K\|4K` | Image size for Google/OpenRouter (default: from quality) |
 | `--imageApiDialect openai-native\|ratio-metadata` | OpenAI-compatible endpoint dialect — use `ratio-metadata` for gateways that expect aspect-ratio `size` plus `metadata.resolution` |
-| `--ref <files...>` | Reference images. Supported by Google multimodal, OpenAI GPT Image edits, Azure OpenAI edits (PNG/JPG only), OpenRouter multimodal models, Replicate supported families, MiniMax subject-reference, Seedream 5.0/4.5/4.0. Not supported by Jimeng, Seedream 3.0, SeedEdit 3.0 |
+| `--ref <files...>` | Reference images. Supported by Google multimodal, OpenAI GPT Image edits, Azure OpenAI edits (PNG/JPG only), OpenRouter multimodal models, Replicate supported families, MiniMax subject-reference, Seedream 5.0/4.5/4.0, DashScope `wan2.7-image-pro`/`wan2.7-image`. Not supported by Jimeng, Seedream 3.0, SeedEdit 3.0, or any DashScope model outside the `wan2.7-image*` family |
 | `--n <count>` | Number of images. Replicate requires `--n 1` (single-output save semantics) |
 | `--json` | JSON output |

@@ -128,7 +131,9 @@ Priority (highest → lowest) applies to every provider:
 3. Env var `<PROVIDER>_IMAGE_MODEL`
 4. Built-in default

-For Azure, `--model` / `default_model.azure` is the Azure deployment name. `AZURE_OPENAI_DEPLOYMENT` is the preferred env var; `AZURE_OPENAI_IMAGE_MODEL` is kept as a backward-compatible alias.
+For OpenAI, the built-in default is `gpt-image-2`. `gpt-image-1.5`, `gpt-image-1`, and GPT Image snapshots remain selectable with `--model` or `OPENAI_IMAGE_MODEL`.
+
+For Azure, `--model` / `default_model.azure` is the Azure deployment name. `AZURE_OPENAI_DEPLOYMENT` is the preferred env var; `AZURE_OPENAI_IMAGE_MODEL` is kept as a backward-compatible alias. If your Azure deployment is named after the underlying model, use `gpt-image-2`; otherwise use the exact custom deployment name.

 EXTEND.md overrides env vars: if EXTEND.md sets `default_model.google: "gemini-3-pro-image-preview"` and the env var sets `GOOGLE_IMAGE_MODEL=gemini-3.1-flash-image-preview`, EXTEND.md wins.

@@ -169,17 +174,19 @@ Each provider has its own quirks (model families, size rules, ref support, limit

 | Preset | Google imageSize | OpenAI size | OpenRouter size | Replicate resolution | Use case |
 |--------|------------------|-------------|-----------------|----------------------|----------|
-| `normal` | 1K | 1024px | 1K | 1K | Quick previews |
-| `2k` (default) | 2K | 2048px | 2K | 2K | Covers, illustrations, infographics |
+| `normal` | 1K | 1024px target | 1K | 1K | Quick previews |
+| `2k` (default) | 2K | 2048px target | 2K | 2K | Covers, illustrations, infographics |

 Google/OpenRouter `imageSize` can be overridden with `--imageSize 1K|2K|4K`.

+For OpenAI native `gpt-image-2`, `normal` maps to `quality=medium` and a low-latency valid size near the requested aspect ratio; `2k` maps to `quality=high` and 2048px-class sizes such as `2048x2048`, `2048x1152`, or `1152x2048`. Use explicit `--size` for valid custom or 4K outputs, e.g. `3840x2160`.
+
 ## Aspect Ratios

 Supported: `1:1`, `16:9`, `9:16`, `4:3`, `3:4`, `2.35:1`.

 - Google multimodal: `imageConfig.aspectRatio`
- OpenAI: closest supported size
+- OpenAI: `gpt-image-2` uses the closest valid custom size for the requested ratio; older GPT Image and DALL·E models use their closest supported fixed size
 - OpenRouter: `imageGenerationOptions.aspect_ratio`; if only `--size <WxH>` is given, the ratio is inferred
 - Replicate: behavior is model-specific — `google/nano-banana*` uses `aspect_ratio`, `bytedance/seedream-*` uses documented Replicate ratios, Wan 2.7 maps `--ar` to a concrete `size`
 - MiniMax: official `aspect_ratio` values; if `--size <WxH>` is given without `--ar`, sends `width`/`height` for `image-01`
@@ -46,7 +46,7 @@ options:
  - label: "Google (Recommended)"
    description: "Gemini multimodal - high quality, reference images, flexible sizes"
  - label: "OpenAI"
-    description: "GPT Image - consistent quality, reliable output"
+    description: "GPT Image 2 - latest OpenAI image model, reference-image workflows"
  - label: "Azure OpenAI"
    description: "Azure-hosted GPT Image deployments with resource-specific routing"
  - label: "OpenRouter"
@@ -101,10 +101,12 @@ Only show if user selected Azure OpenAI.
 header: "Azure Deploy"
 question: "Default Azure image deployment name?"
 options:
-  - label: "gpt-image-1.5 (Recommended)"
-    description: "Best default if your Azure deployment uses the same name"
-  - label: "gpt-image-1"
+  - label: "gpt-image-2 (Recommended)"
+    description: "Use if your Azure deployment uses the GPT Image 2 model name"
+  - label: "gpt-image-1.5"
    description: "Previous GPT Image deployment name"
+  - label: "gpt-image-1"
+    description: "Earlier GPT Image deployment name"
 ```

 ### Question 2d: Default MiniMax Model
@@ -214,10 +216,12 @@ options:
 header: "OpenAI Model"
 question: "Choose a default OpenAI image generation model?"
 options:
-  - label: "gpt-image-1.5 (Recommended)"
-    description: "Latest GPT Image model, high quality"
+  - label: "gpt-image-2 (Recommended)"
+    description: "Latest GPT Image model, flexible sizes up to 4K, high-fidelity image inputs"
+  - label: "gpt-image-1.5"
+    description: "Previous GPT Image model"
  - label: "gpt-image-1"
-    description: "Previous generation GPT Image model"
+    description: "Earlier GPT Image model"
 ```

 ### Azure Deployment Selection
@@ -226,8 +230,10 @@ options:
 header: "Azure Deploy"
 question: "Choose a default Azure image deployment name?"
 options:
-  - label: "gpt-image-1.5 (Recommended)"
-    description: "Use when your Azure deployment name matches the GPT-image-1.5 model"
+  - label: "gpt-image-2 (Recommended)"
+    description: "Use when your Azure deployment name matches the GPT Image 2 model"
+  - label: "gpt-image-1.5"
+    description: "Use when your Azure deployment name matches the GPT Image 1.5 model"
  - label: "gpt-image-1"
    description: "Use when your Azure deployment name matches GPT-image-1"
 ```
@@ -265,6 +271,10 @@ options:
    description: "Legacy Qwen model with five fixed output sizes"
  - label: "qwen-image-plus"
    description: "Legacy Qwen model, same current capability as qwen-image"
+  - label: "wan2.7-image-pro"
+    description: "Wan 2.7 Pro — supports up to 4K text-to-image and reference-image editing"
+  - label: "wan2.7-image"
+    description: "Wan 2.7 base — faster generation, up to 2K, supports reference-image editing"
  - label: "z-image-turbo"
    description: "Legacy DashScope model for compatibility"
  - label: "z-image-ultra"
@@ -275,6 +285,7 @@ Notes for DashScope setup:

 - Prefer `qwen-image-2.0-pro` when the user needs custom `--size`, uncommon ratios like `21:9`, or strong Chinese/English text rendering.
 - `qwen-image-max` / `qwen-image-plus` / `qwen-image` only support five fixed sizes: `1664*928`, `1472*1104`, `1328*1328`, `1104*1472`, `928*1664`.
+- `wan2.7-image-pro` and `wan2.7-image` are the only DashScope models that accept `--ref`. Pick one of these when the user wants reference-image editing or multi-image fusion via DashScope.
 - In `baoyu-imagine`, `quality` is a compatibility preset. It is not a native DashScope parameter.

 ### Z.AI Model Selection
@@ -23,8 +23,8 @@ default_image_api_dialect: null  # openai-native|ratio-metadata|null (OpenAI-com

 default_model:
  google: null              # e.g., "gemini-3-pro-image-preview", "gemini-3.1-flash-image-preview"
-  openai: null              # e.g., "gpt-image-1.5", "gpt-image-1"
-  azure: null               # Azure deployment name, e.g., "gpt-image-1.5" or "image-prod"
+  openai: null              # e.g., "gpt-image-2", "gpt-image-1.5", "gpt-image-1"
+  azure: null               # Azure deployment name, e.g., "gpt-image-2" or "image-prod"
  openrouter: null          # e.g., "google/gemini-3.1-flash-image-preview"
  dashscope: null           # e.g., "qwen-image-2.0-pro"
  zai: null                 # e.g., "glm-image"
@@ -106,8 +106,8 @@ default_image_size: 2K
 default_image_api_dialect: null
 default_model:
  google: "gemini-3-pro-image-preview"
-  openai: "gpt-image-1.5"
-  azure: "gpt-image-1.5"
+  openai: "gpt-image-2"
+  azure: "gpt-image-2"
  openrouter: "google/gemini-3.1-flash-image-preview"
  dashscope: "qwen-image-2.0-pro"
  zai: "glm-image"
@@ -17,6 +17,17 @@ Read when the user picks `--provider dashscope`, sets `default_model.dashscope`,
 - Default is `1664*928`
 - `qwen-image` currently has the same capability as `qwen-image-plus`

+**`wan2.7-image*`** — multimodal Wan 2.7 family. Members: `wan2.7-image-pro`, `wan2.7-image`.
+
+- Free-form `size` in `宽*高` format, plus aspect-ratio inference
+- `wan2.7-image-pro` text-to-image (no `--ref`): total pixels in `[768*768, 4096*4096]`, ratio in `[1:8, 8:1]`
+- `wan2.7-image-pro` with reference images and `wan2.7-image` (all scenarios): total pixels in `[768*768, 2048*2048]`, ratio in `[1:8, 8:1]`
+- Default: `1024*1024` (`--quality normal`) or `2048*2048` (`--quality 2k`); 4K requires explicit `--size`
+- Supports up to 9 reference images in `--ref` (image editing / multi-image fusion)
+- Reference images are sent inline as base64 (or passed through if the path is an `http(s)://` URL)
+- API does NOT use `prompt_extend`; the skill omits it for this family
+- The Wan 2.7 API defaults `n` to **4** in non-collage mode and bills per generated image. baoyu-imagine forces `n: 1` and rejects `--n > 1` to avoid silently paying for and discarding extra images.
+
 **Legacy** — `z-image-turbo`, `z-image-ultra`, `wanx-v1`. Only use when the user explicitly asks for legacy behavior.

 ## Size Resolution
@@ -24,7 +35,8 @@ Read when the user picks `--provider dashscope`, sets `default_model.dashscope`,
 - `--size` wins over `--ar`
 - For `qwen-image-2.0*`: prefer explicit `--size`; otherwise infer from `--ar` using the recommended table below
 - For `qwen-image-max/plus/image`: only use the five fixed sizes; if the requested ratio doesn't fit, switch to `qwen-image-2.0-pro`
- `--quality` is a baoyu-imagine preset, not an official DashScope field. The mapping of `normal`/`2k` onto the `qwen-image-2.0*` table is an implementation choice, not an API guarantee
+- For `wan2.7-image*`: explicit `--size` is validated against the per-mode pixel/ratio limits; otherwise the size is derived from `--ar` and `--quality` (`normal` ≈ 1K, `2k` ≈ 2K). To request 4K with `wan2.7-image-pro` text-to-image, pass `--size` explicitly (e.g. `4096*4096`, `3840*2160`)
+- `--quality` is a baoyu-imagine preset, not an official DashScope field. The mapping of `normal`/`2k` onto the `qwen-image-2.0*` and `wan2.7-image*` tables is an implementation choice, not an API guarantee

 ### Recommended `qwen-image-2.0*` sizes

@@ -39,12 +51,19 @@ Read when the user picks `--provider dashscope`, sets `default_model.dashscope`,
 | `16:9` | `1280*720` | `1920*1080` |
 | `21:9` | `1344*576` | `2048*872` |

+## Reference Images
+
+- Only `wan2.7-image-pro` and `wan2.7-image` accept `--ref`. Other DashScope models (qwen-image-2.0*, qwen-image-max/plus/image, legacy) reject `--ref` and the user is steered to a different provider/model.
+- Up to 9 reference images per request. Local files are inlined as base64 data URLs; `http(s)://` URLs are forwarded as-is.
+- Supplying any `--ref` automatically clamps the wan2.7-image-pro pixel ceiling from 4K to 2K (the API only supports 4K for pure text-to-image with no image input).
+
 ## Not Exposed

-DashScope APIs also support `negative_prompt`, `prompt_extend`, and `watermark`, but `baoyu-imagine` does not expose them as CLI flags today.
+DashScope APIs also support `negative_prompt`, `prompt_extend`, `watermark`, `thinking_mode`, `seed`, `bbox_list`, `enable_sequential`, and `color_palette`. `baoyu-imagine` does not expose them as CLI flags today; the wan2.7 family relies on the API defaults (e.g. `thinking_mode=true`). The skill always sends `n=1` for wan2.7 — if you want grid/collage mode you currently need to call the API directly.

 ## Official References

 - [Qwen-Image API](https://help.aliyun.com/zh/model-studio/qwen-image-api)
 - [Text-to-image guide](https://help.aliyun.com/zh/model-studio/text-to-image)
 - [Qwen-Image Edit API](https://help.aliyun.com/zh/model-studio/qwen-image-edit-api)
+- [Wan 2.7 image generation & editing API](https://help.aliyun.com/zh/model-studio/wan-image-generation-and-editing-api-reference)
@@ -25,10 +25,13 @@ ${BUN_X} {baseDir}/scripts/main.ts --prompt "Make blue" --image out.png --ref so

 ```bash
 # OpenAI
-${BUN_X} {baseDir}/scripts/main.ts --prompt "A cat" --image out.png --provider openai
+${BUN_X} {baseDir}/scripts/main.ts --prompt "A cat" --image out.png --provider openai --model gpt-image-2

 # Azure OpenAI (model = deployment name)
-${BUN_X} {baseDir}/scripts/main.ts --prompt "A cat" --image out.png --provider azure --model gpt-image-1.5
+${BUN_X} {baseDir}/scripts/main.ts --prompt "A cat" --image out.png --provider azure --model gpt-image-2
+
+# OpenAI GPT Image 2 custom 4K size
+${BUN_X} {baseDir}/scripts/main.ts --prompt "A cinematic landscape" --image out.png --provider openai --model gpt-image-2 --size 3840x2160

 # Google with explicit model
 ${BUN_X} {baseDir}/scripts/main.ts --prompt "Make blue" --image out.png --provider google --model gemini-3-pro-image-preview --ref source.png
@@ -48,6 +51,12 @@ ${BUN_X} {baseDir}/scripts/main.ts --prompt "为咖啡品牌设计一张 21:9
 # DashScope legacy fixed-size
 ${BUN_X} {baseDir}/scripts/main.ts --prompt "一张电影感海报" --image out.png --provider dashscope --model qwen-image-max --size 1664x928

+# DashScope Wan 2.7 Image Pro (4K text-to-image)
+${BUN_X} {baseDir}/scripts/main.ts --prompt "一间有着精致窗户的花店" --image out.png --provider dashscope --model wan2.7-image-pro --size 4096x4096
+
+# DashScope Wan 2.7 Image with reference image (multi-image fusion)
+${BUN_X} {baseDir}/scripts/main.ts --prompt "把图2的涂鸦喷绘在图1的汽车上" --image out.png --provider dashscope --model wan2.7-image-pro --ref car.webp paint.webp
+
 # Z.AI GLM-image
 ${BUN_X} {baseDir}/scripts/main.ts --prompt "一张带清晰中文标题的科技海报" --image out.png --provider zai

@@ -19,6 +19,7 @@ import {
  parseArgs,
  parseOpenAIImageApiDialect,
  parseSimpleYaml,
+  validateReferenceImages,
 } from "./main.ts";

 function makeArgs(overrides: Partial<CliArgs> = {}): CliArgs {
@@ -123,6 +124,15 @@ test("parseArgs falls back to positional prompt and rejects invalid provider", (
  );
 });

+test("validateReferenceImages can skip remote URLs for providers that support them", async () => {
+  await validateReferenceImages(["https://example.com/ref.png"], { allowRemoteUrls: true });
+
+  await assert.rejects(
+    () => validateReferenceImages(["https://example.com/ref.png"]),
+    /Reference image not found/,
+  );
+});
+
 test("parseSimpleYaml parses nested defaults and provider limits", () => {
  const yaml = `
 version: 2
@@ -133,7 +143,7 @@ default_image_size: 2K
 default_image_api_dialect: ratio-metadata
 default_model:
  google: gemini-3-pro-image-preview
-  openai: gpt-image-1.5
+  openai: gpt-image-2
  zai: glm-image
  azure: image-prod
  minimax: image-01
@@ -165,7 +175,7 @@ batch:
  assert.equal(config.default_image_size, "2K");
  assert.equal(config.default_image_api_dialect, "ratio-metadata");
  assert.equal(config.default_model?.google, "gemini-3-pro-image-preview");
-  assert.equal(config.default_model?.openai, "gpt-image-1.5");
+  assert.equal(config.default_model?.openai, "gpt-image-2");
  assert.equal(config.default_model?.zai, "glm-image");
  assert.equal(config.default_model?.azure, "image-prod");
  assert.equal(config.default_model?.minimax, "image-01");
@@ -308,7 +318,7 @@ test("detectProvider rejects non-ref-capable providers and prefers Google first
    () =>
      detectProvider(
        makeArgs({
-          provider: "dashscope",
+          provider: "zai",
          referenceImages: ["ref.png"],
        }),
      ),
@@ -426,6 +436,33 @@ test("detectProvider infers Seedream from model id and allows Seedream reference
  );
 });

+test("detectProvider allows DashScope reference-image workflows when explicitly chosen for wan2.7 models", (t) => {
+  useEnv(t, {
+    GOOGLE_API_KEY: null,
+    OPENAI_API_KEY: null,
+    AZURE_OPENAI_API_KEY: null,
+    AZURE_OPENAI_BASE_URL: null,
+    OPENROUTER_API_KEY: null,
+    DASHSCOPE_API_KEY: "dashscope-key",
+    MINIMAX_API_KEY: null,
+    REPLICATE_API_TOKEN: null,
+    JIMENG_ACCESS_KEY_ID: null,
+    JIMENG_SECRET_ACCESS_KEY: null,
+    ARK_API_KEY: null,
+  });
+
+  assert.equal(
+    detectProvider(
+      makeArgs({
+        provider: "dashscope",
+        model: "wan2.7-image-pro",
+        referenceImages: ["ref.png"],
+      }),
+    ),
+    "dashscope",
+  );
+});
+
 test("detectProvider selects MiniMax when only MiniMax credentials are configured or the model id matches", (t) => {
  useEnv(t, {
    GOOGLE_API_KEY: null,
@@ -504,7 +541,7 @@ test("loadBatchTasks and createTaskArgs resolve batch-relative paths", async (t)
          id: "hero",
          promptFiles: ["prompts/hero.md"],
          image: "out/hero",
-          ref: ["refs/hero.png"],
+          ref: ["refs/hero.png", "https://example.com/ref.png"],
          ar: "16:9",
        },
      ],
@@ -533,6 +570,7 @@ test("loadBatchTasks and createTaskArgs resolve batch-relative paths", async (t)
  assert.equal(taskArgs.imagePath, path.join(loaded.batchDir, "out/hero"));
  assert.deepEqual(taskArgs.referenceImages, [
    path.join(loaded.batchDir, "refs/hero.png"),
+    "https://example.com/ref.png",
  ]);
  assert.equal(taskArgs.provider, "replicate");
  assert.equal(taskArgs.aspectRatio, "16:9");
@@ -557,5 +595,29 @@ test("path normalization, worker count, and retry classification follow expected
    ),
    false,
  );
+  assert.equal(
+    isRetryableGenerationError(
+      new Error("DashScope wan2.7 image models accept at most 9 reference images. Received 10."),
+    ),
+    false,
+  );
+  assert.equal(
+    isRetryableGenerationError(
+      new Error("DashScope wan2.7 image models in baoyu-imagine support exactly one output image per request."),
+    ),
+    false,
+  );
+  assert.equal(
+    isRetryableGenerationError(
+      new Error("DashScope wan2.7 image models support aspect ratios in [1:8, 8:1]."),
+    ),
+    false,
+  );
+  assert.equal(
+    isRetryableGenerationError(
+      new Error("DashScope wan2.7-image requires total pixels between 768*768 and 2048*2048."),
+    ),
+    false,
+  );
  assert.equal(isRetryableGenerationError(new Error("socket hang up")), true);
 });
@@ -85,7 +85,7 @@ Options:
  --quality normal|2k       Quality preset (default: 2k)
  --imageSize 1K|2K|4K      Image size for Google/OpenRouter (default: from quality)
  --imageApiDialect <id>    OpenAI-compatible image dialect: openai-native|ratio-metadata
-  --ref <files...>          Reference images (Google, OpenAI, Azure, OpenRouter, Replicate supported families, MiniMax, or Seedream 4.0/4.5/5.0)
+  --ref <files...>          Reference images (Google, OpenAI, Azure, OpenRouter, Replicate supported families, MiniMax, Seedream 4.0/4.5/5.0, or DashScope wan2.7-image*)
  --n <count>               Number of images for the current task (default: 1; Replicate currently requires 1)
  --json                    JSON output
  -h, --help                Show help
@@ -124,7 +124,7 @@ Environment variables:
  JIMENG_ACCESS_KEY_ID      Jimeng Access Key ID
  JIMENG_SECRET_ACCESS_KEY  Jimeng Secret Access Key
  ARK_API_KEY               Seedream/Ark API key
-  OPENAI_IMAGE_MODEL        Default OpenAI model (gpt-image-1.5)
+  OPENAI_IMAGE_MODEL        Default OpenAI model (gpt-image-2)
  OPENROUTER_IMAGE_MODEL    Default OpenRouter model (google/gemini-3.1-flash-image-preview)
  GOOGLE_IMAGE_MODEL        Default Google model (gemini-3-pro-image-preview)
  DASHSCOPE_IMAGE_MODEL     Default DashScope model (qwen-image-2.0-pro)
@@ -151,7 +151,7 @@ Environment variables:
  AZURE_OPENAI_BASE_URL     Azure OpenAI resource or deployment endpoint
  AZURE_OPENAI_DEPLOYMENT   Default Azure deployment name
  AZURE_API_VERSION         Azure API version (default: 2025-04-01-preview)
-  AZURE_OPENAI_IMAGE_MODEL  Backward-compatible Azure deployment/model alias (defaults to gpt-image-1.5)
+  AZURE_OPENAI_IMAGE_MODEL  Backward-compatible Azure deployment/model alias (defaults to gpt-image-2)
  SEEDREAM_BASE_URL         Custom Seedream endpoint
  BAOYU_IMAGE_GEN_MAX_WORKERS  Override batch worker cap
  BAOYU_IMAGE_GEN_<PROVIDER>_CONCURRENCY  Override provider concurrency
@@ -698,10 +698,11 @@ export function detectProvider(args: CliArgs): Provider {
    args.provider !== "openrouter" &&
    args.provider !== "replicate" &&
    args.provider !== "seedream" &&
-    args.provider !== "minimax"
+    args.provider !== "minimax" &&
+    args.provider !== "dashscope"
  ) {
    throw new Error(
-      "Reference images require a ref-capable provider. Use --provider google (Gemini multimodal), --provider openai (GPT Image edits), --provider azure (Azure OpenAI), --provider openrouter (OpenRouter multimodal), --provider replicate, --provider seedream for supported Seedream models, or --provider minimax for MiniMax subject-reference workflows."
+      "Reference images require a ref-capable provider. Use --provider google (Gemini multimodal), --provider openai (GPT Image edits), --provider azure (Azure OpenAI), --provider openrouter (OpenRouter multimodal), --provider replicate, --provider dashscope with a wan2.7 image model, --provider seedream for supported Seedream models, or --provider minimax for MiniMax subject-reference workflows."
    );
  }

@@ -775,8 +776,24 @@ export function detectProvider(args: CliArgs): Provider {
  );
 }

-export async function validateReferenceImages(referenceImages: string[]): Promise<void> {
+export type ReferenceImageValidationOptions = {
+  allowRemoteUrls?: boolean;
+};
+
+function isRemoteReferenceImage(refPath: string): boolean {
+  return /^https?:\/\//i.test(refPath);
+}
+
+function shouldAllowRemoteReferenceImages(provider: Provider | null): boolean {
+  return provider === "dashscope";
+}
+
+export async function validateReferenceImages(
+  referenceImages: string[],
+  options: ReferenceImageValidationOptions = {},
+): Promise<void> {
  for (const refPath of referenceImages) {
+    if (options.allowRemoteUrls && isRemoteReferenceImage(refPath)) continue;
    const fullPath = path.resolve(refPath);
    try {
      await access(fullPath);
@@ -803,6 +820,11 @@ export function isRetryableGenerationError(error: unknown): boolean {
    "API error (404)",
    "temporarily disabled",
    "supports saving exactly one image",
+    "supports only",
+    "support exactly one output image",
+    "support aspect ratios in",
+    "requires total pixels between",
+    "accept at most",
  ];
  return !nonRetryableMarkers.some((marker) => msg.includes(marker));
 }
@@ -858,7 +880,11 @@ async function prepareSingleTask(args: CliArgs, extendConfig: Partial<ExtendConf
  const prompt = (await loadPromptForArgs(args)) ?? (await readPromptFromStdin());
  if (!prompt) throw new Error("Prompt is required");
  if (!args.imagePath) throw new Error("--image is required");
-  if (args.referenceImages.length > 0) await validateReferenceImages(args.referenceImages);
+  if (args.referenceImages.length > 0) {
+    await validateReferenceImages(args.referenceImages, {
+      allowRemoteUrls: shouldAllowRemoteReferenceImages(args.provider),
+    });
+  }

  const provider = detectProvider(args);
  const providerModule = await loadProviderModule(provider);
@@ -907,6 +933,10 @@ export function resolveBatchPath(batchDir: string, filePath: string): string {
  return path.isAbsolute(filePath) ? filePath : path.resolve(batchDir, filePath);
 }

+function resolveBatchReferencePath(batchDir: string, filePath: string): string {
+  return isRemoteReferenceImage(filePath) ? filePath : resolveBatchPath(batchDir, filePath);
+}
+
 export function createTaskArgs(baseArgs: CliArgs, task: BatchTaskInput, batchDir: string): CliArgs {
  return {
    ...baseArgs,
@@ -922,7 +952,7 @@ export function createTaskArgs(baseArgs: CliArgs, task: BatchTaskInput, batchDir
    imageSize: task.imageSize ?? baseArgs.imageSize ?? null,
    imageSizeSource: task.imageSize != null ? "task" : (baseArgs.imageSizeSource ?? null),
    imageApiDialect: task.imageApiDialect ?? baseArgs.imageApiDialect ?? null,
-    referenceImages: task.ref ? task.ref.map((filePath) => resolveBatchPath(batchDir, filePath)) : [],
+    referenceImages: task.ref ? task.ref.map((filePath) => resolveBatchReferencePath(batchDir, filePath)) : [],
    n: task.n ?? baseArgs.n,
    batchFile: null,
    jobs: baseArgs.jobs,
@@ -946,7 +976,11 @@ async function prepareBatchTasks(
    const prompt = await loadPromptForArgs(taskArgs);
    if (!prompt) throw new Error(`Task ${i + 1} is missing prompt or promptFiles.`);
    if (!taskArgs.imagePath) throw new Error(`Task ${i + 1} is missing image output path.`);
-    if (taskArgs.referenceImages.length > 0) await validateReferenceImages(taskArgs.referenceImages);
+    if (taskArgs.referenceImages.length > 0) {
+      await validateReferenceImages(taskArgs.referenceImages, {
+        allowRemoteUrls: shouldAllowRemoteReferenceImages(taskArgs.provider),
+      });
+    }

    const provider = detectProvider(taskArgs);
    const providerModule = await loadProviderModule(provider);
@@ -46,7 +46,7 @@ export function getDefaultModel(): string {
    }
  }

-  return process.env.AZURE_OPENAI_IMAGE_MODEL || "gpt-image-1.5";
+  return process.env.AZURE_OPENAI_IMAGE_MODEL || "gpt-image-2";
 }

 function getEndpoint(): AzureEndpoint {
@@ -2,15 +2,42 @@ import assert from "node:assert/strict";
 import test, { type TestContext } from "node:test";

 import {
+  generateImage,
  getDefaultModel,
  getModelFamily,
  getQwen2SizeFromAspectRatio,
  getSizeFromAspectRatio,
+  getWan27SizeFromAspectRatio,
  normalizeSize,
  parseAspectRatio,
  parseSize,
  resolveSizeForModel,
 } from "./dashscope.ts";
+import type { CliArgs } from "../types.ts";
+
+function makeCliArgs(overrides: Partial<CliArgs> = {}): CliArgs {
+  return {
+    prompt: null,
+    promptFiles: [],
+    imagePath: null,
+    provider: "dashscope",
+    model: null,
+    aspectRatio: null,
+    aspectRatioSource: null,
+    size: null,
+    quality: "2k",
+    imageSize: null,
+    imageSizeSource: null,
+    imageApiDialect: null,
+    referenceImages: [],
+    n: 1,
+    batchFile: null,
+    jobs: null,
+    json: false,
+    help: false,
+    ...overrides,
+  };
+}

 function useEnv(
  t: TestContext,
@@ -51,9 +78,11 @@ test("DashScope aspect-ratio parsing accepts numeric ratios only", () => {
  assert.equal(parseAspectRatio("-1:2"), null);
 });

-test("DashScope model family routing distinguishes qwen-2.0, fixed-size qwen, and legacy models", () => {
+test("DashScope model family routing distinguishes qwen-2.0, fixed-size qwen, wan2.7, and legacy models", () => {
  assert.equal(getModelFamily("qwen-image-2.0-pro"), "qwen2");
  assert.equal(getModelFamily("qwen-image"), "qwenFixed");
+  assert.equal(getModelFamily("wan2.7-image"), "wan27");
+  assert.equal(getModelFamily("wan2.7-image-pro"), "wan27");
  assert.equal(getModelFamily("z-image-turbo"), "legacy");
  assert.equal(getModelFamily("wanx-v1"), "legacy");
 });
@@ -146,3 +175,218 @@ test("DashScope size normalization converts WxH into provider format", () => {
  assert.equal(normalizeSize("1024x1024"), "1024*1024");
  assert.equal(normalizeSize("2048*1152"), "2048*1152");
 });
+
+test("Wan 2.7 derives sizes that match the requested ratio at the chosen pixel budget", () => {
+  const square2k = getWan27SizeFromAspectRatio(null, "2k", 2048 * 2048);
+  const parsedSquare = parseSize(square2k);
+  assert.ok(parsedSquare);
+  assert.equal(parsedSquare.width, parsedSquare.height);
+  assert.ok(parsedSquare.width * parsedSquare.height <= 2048 * 2048);
+
+  const widescreen = getWan27SizeFromAspectRatio("16:9", "2k", 2048 * 2048);
+  const parsedWide = parseSize(widescreen);
+  assert.ok(parsedWide);
+  assert.ok(Math.abs(parsedWide.width / parsedWide.height - 16 / 9) < 0.05);
+  assert.ok(parsedWide.width * parsedWide.height <= 2048 * 2048);
+
+  const pro4k = getWan27SizeFromAspectRatio("16:9", "2k", 4096 * 4096);
+  const parsed4k = parseSize(pro4k);
+  assert.ok(parsed4k);
+  assert.ok(parsed4k.width * parsed4k.height > 2048 * 2048);
+  assert.ok(parsed4k.width * parsed4k.height <= 4096 * 4096);
+});
+
+test("Wan 2.7 rejects aspect ratios outside the [1:8, 8:1] range", () => {
+  assert.throws(
+    () => getWan27SizeFromAspectRatio("9:1", "2k", 2048 * 2048),
+    /1:8, 8:1/,
+  );
+  assert.throws(
+    () => getWan27SizeFromAspectRatio("1:9", "normal", 2048 * 2048),
+    /1:8, 8:1/,
+  );
+});
+
+test("Wan 2.7 derived sizes stay inside the boundary ratio limits after rounding", () => {
+  for (const ar of ["8:1", "1:8"]) {
+    const size = getWan27SizeFromAspectRatio(ar, "2k", 2048 * 2048);
+    const parsed = parseSize(size);
+    assert.ok(parsed);
+    const ratio = parsed.width / parsed.height;
+    assert.ok(ratio >= 1 / 8);
+    assert.ok(ratio <= 8);
+    assert.ok(parsed.width * parsed.height <= 2048 * 2048);
+  }
+});
+
+test("resolveSizeForModel routes wan2.7-image to the 2K-capped derivation", () => {
+  const size = resolveSizeForModel("wan2.7-image", {
+    size: null,
+    aspectRatio: "16:9",
+    quality: "2k",
+  });
+  const parsed = parseSize(size);
+  assert.ok(parsed);
+  assert.ok(parsed.width * parsed.height <= 2048 * 2048);
+  assert.ok(Math.abs(parsed.width / parsed.height - 16 / 9) < 0.05);
+});
+
+test("resolveSizeForModel allows wan2.7-image-pro 4K only when there are no reference images", () => {
+  assert.equal(
+    resolveSizeForModel("wan2.7-image-pro", {
+      size: "4096*4096",
+      aspectRatio: null,
+      quality: "2k",
+    }),
+    "4096*4096",
+  );
+
+  assert.throws(
+    () =>
+      resolveSizeForModel("wan2.7-image-pro", {
+        size: "4096*4096",
+        aspectRatio: null,
+        quality: "2k",
+        referenceImages: ["a.png"],
+      }),
+    /total pixels between 768\*768 and 2048\*2048/,
+  );
+
+  const proWithRef = resolveSizeForModel("wan2.7-image-pro", {
+    size: null,
+    aspectRatio: "1:1",
+    quality: "2k",
+    referenceImages: ["a.png"],
+  });
+  const parsedRef = parseSize(proWithRef);
+  assert.ok(parsedRef);
+  assert.ok(parsedRef.width * parsedRef.height <= 2048 * 2048);
+});
+
+test("Wan 2.7 request body forces n=1 and omits prompt_extend / negative_prompt", async (t) => {
+  useEnv(t, { DASHSCOPE_API_KEY: "fake-key" });
+
+  const originalFetch = globalThis.fetch;
+  let capturedBody: any = null;
+  globalThis.fetch = (async (_url: string, init?: RequestInit) => {
+    capturedBody = JSON.parse(String(init?.body));
+    return new Response(
+      JSON.stringify({
+        output: {
+          choices: [
+            {
+              message: {
+                content: [{ image: "data:image/png;base64,iVBORw0KGgo=" }],
+              },
+            },
+          ],
+        },
+      }),
+      { status: 200, headers: { "content-type": "application/json" } },
+    );
+  }) as typeof fetch;
+  t.after(() => {
+    globalThis.fetch = originalFetch;
+  });
+
+  await generateImage("hello", "wan2.7-image-pro", makeCliArgs({ aspectRatio: "1:1" }));
+
+  assert.equal(capturedBody.model, "wan2.7-image-pro");
+  assert.deepEqual(Object.keys(capturedBody.parameters).sort(), ["n", "size", "watermark"]);
+  assert.equal(capturedBody.parameters.n, 1);
+  assert.equal(capturedBody.parameters.watermark, false);
+  assert.equal(typeof capturedBody.parameters.size, "string");
+  assert.ok(!("prompt_extend" in capturedBody.parameters));
+  assert.ok(!("negative_prompt" in capturedBody.parameters));
+
+  assert.deepEqual(capturedBody.input.messages[0].content, [{ text: "hello" }]);
+});
+
+test("Wan 2.7 request body forwards remote reference image URLs", async (t) => {
+  useEnv(t, { DASHSCOPE_API_KEY: "fake-key" });
+
+  const originalFetch = globalThis.fetch;
+  let capturedBody: any = null;
+  globalThis.fetch = (async (_url: string, init?: RequestInit) => {
+    capturedBody = JSON.parse(String(init?.body));
+    return new Response(
+      JSON.stringify({
+        output: {
+          choices: [
+            {
+              message: {
+                content: [{ image: "data:image/png;base64,iVBORw0KGgo=" }],
+              },
+            },
+          ],
+        },
+      }),
+      { status: 200, headers: { "content-type": "application/json" } },
+    );
+  }) as typeof fetch;
+  t.after(() => {
+    globalThis.fetch = originalFetch;
+  });
+
+  await generateImage(
+    "combine these",
+    "wan2.7-image-pro",
+    makeCliArgs({ referenceImages: ["https://example.com/ref.png"] }),
+  );
+
+  assert.deepEqual(capturedBody.input.messages[0].content, [
+    { image: "https://example.com/ref.png" },
+    { text: "combine these" },
+  ]);
+});
+
+test("Wan 2.7 rejects --n > 1 to prevent silent multi-image billing", async (t) => {
+  useEnv(t, { DASHSCOPE_API_KEY: "fake-key" });
+
+  await assert.rejects(
+    () => generateImage("hi", "wan2.7-image-pro", makeCliArgs({ n: 2 })),
+    /support exactly one output image/,
+  );
+});
+
+test("resolveSizeForModel validates explicit wan2.7 sizes by pixel budget and ratio", () => {
+  assert.equal(
+    resolveSizeForModel("wan2.7-image-pro", {
+      size: "3840x2160",
+      aspectRatio: null,
+      quality: "2k",
+    }),
+    "3840*2160",
+  );
+
+  assert.throws(
+    () =>
+      resolveSizeForModel("wan2.7-image-pro", {
+        size: "3840x2160",
+        aspectRatio: null,
+        quality: "2k",
+        referenceImages: ["a.png"],
+      }),
+    /total pixels between 768\*768 and 2048\*2048/,
+  );
+
+  assert.throws(
+    () =>
+      resolveSizeForModel("wan2.7-image", {
+        size: "4096x4096",
+        aspectRatio: null,
+        quality: "2k",
+      }),
+    /total pixels between 768\*768 and 2048\*2048/,
+  );
+
+  assert.throws(
+    () =>
+      resolveSizeForModel("wan2.7-image-pro", {
+        size: "3072*256",
+        aspectRatio: null,
+        quality: "2k",
+      }),
+    /1:8, 8:1/,
+  );
+});
@@ -1,6 +1,8 @@
+import path from "node:path";
+import { readFile } from "node:fs/promises";
 import type { CliArgs, Quality } from "../types";

-type DashScopeModelFamily = "qwen2" | "qwenFixed" | "legacy";
+type DashScopeModelFamily = "qwen2" | "qwenFixed" | "wan27" | "legacy";

 type DashScopeModelSpec = {
  family: DashScopeModelFamily;
@@ -19,6 +21,16 @@ const QWEN_2_TARGET_PIXELS: Record<Quality, number> = {
  "2k": 1536 * 1536,
 };

+const MIN_WAN27_TOTAL_PIXELS = 768 * 768;
+const MAX_WAN27_PRO_T2I_PIXELS = 4096 * 4096;
+const MAX_WAN27_GENERAL_PIXELS = 2048 * 2048;
+const WAN27_MAX_REFERENCE_IMAGES = 9;
+
+const WAN27_TARGET_PIXELS: Record<Quality, number> = {
+  normal: 1024 * 1024,
+  "2k": 2048 * 2048,
+};
+
 const QWEN_2_RECOMMENDED: Record<string, Record<Quality, string>> = {
  "1:1": { normal: "1024*1024", "2k": "1536*1536" },
  "2:3": { normal: "768*1152", "2k": "1024*1536" },
@@ -73,6 +85,11 @@ const QWEN_FIXED_SPEC: DashScopeModelSpec = {
  defaultSize: QWEN_FIXED_SIZES_BY_RATIO["16:9"],
 };

+const WAN27_SPEC: DashScopeModelSpec = {
+  family: "wan27",
+  defaultSize: "2048*2048",
+};
+
 const LEGACY_SPEC: DashScopeModelSpec = {
  family: "legacy",
  defaultSize: "1536*1536",
@@ -88,12 +105,31 @@ const MODEL_SPEC_ALIASES: Record<string, DashScopeModelSpec> = {
  "qwen-image-plus": QWEN_FIXED_SPEC,
  "qwen-image-plus-2026-01-09": QWEN_FIXED_SPEC,
  "qwen-image": QWEN_FIXED_SPEC,
+  "wan2.7-image-pro": WAN27_SPEC,
+  "wan2.7-image": WAN27_SPEC,
 };

 export function getDefaultModel(): string {
  return process.env.DASHSCOPE_IMAGE_MODEL || DEFAULT_MODEL;
 }

+function getReferenceImageMime(filePath: string): string {
+  const ext = path.extname(filePath).toLowerCase();
+  if (ext === ".jpg" || ext === ".jpeg") return "image/jpeg";
+  if (ext === ".webp") return "image/webp";
+  if (ext === ".bmp") return "image/bmp";
+  return "image/png";
+}
+
+async function loadReferenceImage(refPath: string): Promise<string> {
+  if (/^https?:\/\//i.test(refPath)) {
+    return refPath;
+  }
+  const fullPath = path.resolve(refPath);
+  const bytes = await readFile(fullPath);
+  return `data:${getReferenceImageMime(fullPath)};base64,${bytes.toString("base64")}`;
+}
+
 function getApiKey(): string | null {
  return process.env.DASHSCOPE_API_KEY || null;
 }
@@ -173,6 +209,10 @@ function roundToStep(value: number): number {
  return Math.max(SIZE_STEP, Math.round(value / SIZE_STEP) * SIZE_STEP);
 }

+function floorToStep(value: number): number {
+  return Math.max(SIZE_STEP, Math.floor(value / SIZE_STEP) * SIZE_STEP);
+}
+
 function fitToPixelBudget(
  width: number,
  height: number,
@@ -220,6 +260,21 @@ function fitToPixelBudget(
  return { width: roundedWidth, height: roundedHeight };
 }

+function clampWan27DerivedSizeToRatioBounds(
+  size: { width: number; height: number },
+): { width: number; height: number } {
+  let { width, height } = size;
+  const ratio = width / height;
+
+  if (ratio > 8) {
+    width = floorToStep(height * 8);
+  } else if (ratio < 1 / 8) {
+    height = floorToStep(width * 8);
+  }
+
+  return { width, height };
+}
+
 export function getSizeFromAspectRatio(ar: string | null, quality: CliArgs["quality"]): string {
  const normalizedQuality = normalizeQuality(quality);
  const sizes = normalizedQuality === "2k" ? LEGACY_STANDARD_SIZES_2K : LEGACY_STANDARD_SIZES;
@@ -276,6 +331,77 @@ export function getQwen2SizeFromAspectRatio(ar: string | null, quality: CliArgs[
  return formatSize(fitted.width, fitted.height);
 }

+function isWan27ProModel(model: string): boolean {
+  return model.trim().toLowerCase() === "wan2.7-image-pro";
+}
+
+function getWan27MaxPixels(model: string, hasReferenceImages: boolean): number {
+  if (isWan27ProModel(model) && !hasReferenceImages) {
+    return MAX_WAN27_PRO_T2I_PIXELS;
+  }
+  return MAX_WAN27_GENERAL_PIXELS;
+}
+
+export function getWan27SizeFromAspectRatio(
+  ar: string | null,
+  quality: CliArgs["quality"],
+  maxPixels: number,
+): string {
+  const normalizedQuality = normalizeQuality(quality);
+  const targetPixels = Math.min(WAN27_TARGET_PIXELS[normalizedQuality], maxPixels);
+
+  if (!ar) {
+    const side = roundToStep(Math.sqrt(targetPixels));
+    return formatSize(side, side);
+  }
+
+  const parsed = parseAspectRatio(ar);
+  if (!parsed) {
+    const side = roundToStep(Math.sqrt(targetPixels));
+    return formatSize(side, side);
+  }
+
+  const ratio = parsed.width / parsed.height;
+  if (ratio < 1 / 8 || ratio > 8) {
+    throw new Error(
+      `DashScope wan2.7 image models support aspect ratios in [1:8, 8:1]. Received "${ar}".`
+    );
+  }
+
+  const rawWidth = Math.sqrt(targetPixels * ratio);
+  const rawHeight = Math.sqrt(targetPixels / ratio);
+  const fitted = fitToPixelBudget(
+    rawWidth,
+    rawHeight,
+    MIN_WAN27_TOTAL_PIXELS,
+    maxPixels,
+  );
+  const bounded = clampWan27DerivedSizeToRatioBounds(fitted);
+
+  return formatSize(bounded.width, bounded.height);
+}
+
+function validateWan27Size(size: string, maxPixels: number, model: string): string {
+  const normalized = normalizeSize(size);
+  const parsed = validateSizeFormat(normalized);
+  const totalPixels = parsed.width * parsed.height;
+  if (totalPixels < MIN_WAN27_TOTAL_PIXELS || totalPixels > maxPixels) {
+    const limit = maxPixels === MAX_WAN27_PRO_T2I_PIXELS ? "4096*4096" : "2048*2048";
+    throw new Error(
+      `DashScope ${model} requires total pixels between 768*768 and ${limit} ` +
+      `for the current request. Received ${normalized} (${totalPixels} pixels).`
+    );
+  }
+  const ratio = parsed.width / parsed.height;
+  if (ratio < 1 / 8 || ratio > 8) {
+    throw new Error(
+      `DashScope wan2.7 image models support aspect ratios in [1:8, 8:1]. ` +
+      `Received ${normalized} (ratio ${ratio.toFixed(3)}).`
+    );
+  }
+  return normalized;
+}
+
 function getQwenFixedSizeFromAspectRatio(ar: string | null, quality: CliArgs["quality"]): string {
  if (quality === "normal") {
    console.warn(
@@ -331,9 +457,16 @@ function validateQwenFixedSize(size: string): string {

 export function resolveSizeForModel(
  model: string,
-  args: Pick<CliArgs, "size" | "aspectRatio" | "quality">,
+  args: Pick<CliArgs, "size" | "aspectRatio" | "quality"> & { referenceImages?: string[] },
 ): string {
  const spec = getModelSpec(model);
+  const referenceCount = args.referenceImages?.length ?? 0;
+
+  if (spec.family === "wan27") {
+    const maxPixels = getWan27MaxPixels(model, referenceCount > 0);
+    if (args.size) return validateWan27Size(args.size, maxPixels, model);
+    return getWan27SizeFromAspectRatio(args.aspectRatio, args.quality, maxPixels);
+  }

  if (args.size) {
    if (spec.family === "qwen2") return validateQwen2Size(args.size);
@@ -357,6 +490,14 @@ function buildParameters(
  family: DashScopeModelFamily,
  size: string,
 ): Record<string, unknown> {
+  if (family === "wan27") {
+    return {
+      size,
+      n: 1,
+      watermark: false,
+    };
+  }
+
  const parameters: Record<string, unknown> = {
    prompt_extend: false,
    size,
@@ -419,23 +560,44 @@ export async function generateImage(
  const apiKey = getApiKey();
  if (!apiKey) throw new Error("DASHSCOPE_API_KEY is required");

-  if (args.referenceImages.length > 0) {
+  const spec = getModelSpec(model);
+
+  if (args.referenceImages.length > 0 && spec.family !== "wan27") {
    throw new Error(
-      "Reference images are not supported with DashScope provider in baoyu-imagine. Use --provider google with a Gemini multimodal model."
+      "Reference images are not supported with this DashScope model. Use a wan2.7 image model (--model wan2.7-image-pro or wan2.7-image), or switch to --provider google with a Gemini multimodal model."
+    );
+  }
+
+  if (args.referenceImages.length > WAN27_MAX_REFERENCE_IMAGES) {
+    throw new Error(
+      `DashScope wan2.7 image models accept at most ${WAN27_MAX_REFERENCE_IMAGES} reference images. Received ${args.referenceImages.length}.`
+    );
+  }
+
+  if (spec.family === "wan27" && args.n !== 1) {
+    throw new Error(
+      "DashScope wan2.7 image models in baoyu-imagine support exactly one output image per request (extra images would be billed but discarded). Remove --n or use --n 1."
    );
  }

-  const spec = getModelSpec(model);
  const size = resolveSizeForModel(model, args);
  const url = `${getBaseUrl()}/api/v1/services/aigc/multimodal-generation/generation`;

+  const content: Array<Record<string, unknown>> = [];
+  if (spec.family === "wan27" && args.referenceImages.length > 0) {
+    for (const refPath of args.referenceImages) {
+      content.push({ image: await loadReferenceImage(refPath) });
+    }
+  }
+  content.push({ text: prompt });
+
  const body = {
    model,
    input: {
      messages: [
        {
          role: "user",
-          content: [{ text: prompt }],
+          content,
        },
      ],
    },
@@ -1,9 +1,11 @@
 import assert from "node:assert/strict";
 import test from "node:test";

+import type { CliArgs } from "../types.ts";
 import {
  buildOpenAIGenerationsBody,
  extractImageFromResponse,
+  getDefaultModel,
  getOpenAIAspectRatio,
  getOpenAIImageApiDialect,
  getOpenAIResolution,
@@ -13,9 +15,33 @@ import {
  inferAspectRatioFromSize,
  inferResolutionFromSize,
  parseAspectRatio,
+  validateArgs,
 } from "./openai.ts";

+function makeArgs(overrides: Partial<CliArgs> = {}): CliArgs {
+  return {
+    prompt: null,
+    promptFiles: [],
+    imagePath: null,
+    provider: null,
+    model: null,
+    aspectRatio: null,
+    size: null,
+    quality: "2k",
+    imageSize: null,
+    imageApiDialect: null,
+    referenceImages: [],
+    n: 1,
+    batchFile: null,
+    jobs: null,
+    json: false,
+    help: false,
+    ...overrides,
+  };
+}
+
 test("OpenAI aspect-ratio parsing and size selection match model families", () => {
+  assert.equal(getDefaultModel(), "gpt-image-2");
  assert.deepEqual(parseAspectRatio("16:9"), { width: 16, height: 9 });
  assert.equal(parseAspectRatio("wide"), null);
  assert.equal(parseAspectRatio("0:1"), null);
@@ -25,6 +51,10 @@ test("OpenAI aspect-ratio parsing and size selection match model families", () =
  assert.equal(getOpenAISize("dall-e-2", "16:9", "2k"), "1024x1024");
  assert.equal(getOpenAISize("gpt-image-1.5", "16:9", "2k"), "1536x1024");
  assert.equal(getOpenAISize("gpt-image-1.5", "4:3", "2k"), "1024x1024");
+  assert.equal(getOpenAISize("gpt-image-2", "16:9", "2k"), "2048x1152");
+  assert.equal(getOpenAISize("gpt-image-2", "9:16", "2k"), "1152x2048");
+  assert.equal(getOpenAISize("gpt-image-2", "4:3", "2k"), "2048x1536");
+  assert.equal(getOpenAISize("gpt-image-2", "2.35:1", "normal"), "1248x528");
  assert.equal(inferAspectRatioFromSize("1536x1024"), "3:2");
  assert.equal(inferResolutionFromSize("1536x1024"), "2K");
  assert.equal(getOpenAIAspectRatio({ aspectRatio: null, size: "2048x1152" }), "16:9");
@@ -37,7 +67,7 @@ test("OpenAI aspect-ratio parsing and size selection match model families", () =

 test("OpenAI generations body switches between native and ratio-metadata dialects", () => {
  assert.deepEqual(
-    buildOpenAIGenerationsBody("Draw a skyline", "gpt-image-1.5", {
+    buildOpenAIGenerationsBody("Draw a skyline", "gpt-image-2", {
      aspectRatio: "16:9",
      size: null,
      quality: "2k",
@@ -45,9 +75,10 @@ test("OpenAI generations body switches between native and ratio-metadata dialect
      imageApiDialect: null,
    }),
    {
-      model: "gpt-image-1.5",
+      model: "gpt-image-2",
      prompt: "Draw a skyline",
-      size: "1536x1024",
+      size: "2048x1152",
+      quality: "high",
    },
  );

@@ -90,6 +121,28 @@ test("OpenAI generations body switches between native and ratio-metadata dialect
  );
 });

+test("OpenAI validates gpt-image-2 custom size constraints", () => {
+  assert.doesNotThrow(() =>
+    validateArgs("gpt-image-2", makeArgs({ size: "3840x2160" })),
+  );
+  assert.doesNotThrow(() =>
+    validateArgs("gpt-image-2-2026-04-21", makeArgs({ aspectRatio: "2.35:1" })),
+  );
+
+  assert.throws(
+    () => validateArgs("gpt-image-2", makeArgs({ size: "1024x576" })),
+    /total pixels/,
+  );
+  assert.throws(
+    () => validateArgs("gpt-image-2", makeArgs({ size: "1025x1024" })),
+    /multiples of 16px/,
+  );
+  assert.throws(
+    () => validateArgs("gpt-image-2", makeArgs({ aspectRatio: "4:1" })),
+    /must not exceed 3:1/,
+  );
+});
+
 test("OpenAI mime-type detection covers supported reference image extensions", () => {
  assert.equal(getMimeType("frame.png"), "image/png");
  assert.equal(getMimeType("frame.jpg"), "image/jpeg");
@@ -3,7 +3,7 @@ import { readFile } from "node:fs/promises";
 import type { CliArgs, OpenAIImageApiDialect } from "../types";

 export function getDefaultModel(): string {
-  return process.env.OPENAI_IMAGE_MODEL || "gpt-image-1.5";
+  return process.env.OPENAI_IMAGE_MODEL || "gpt-image-2";
 }

 type OpenAIImageResponse = { data: Array<{ url?: string; b64_json?: string }> };
@@ -25,6 +25,55 @@ type SizeMapping = {

 type OpenAIGenerationsBody = Record<string, unknown>;

+function isGptImageModel(model: string): boolean {
+  return model.includes("gpt-image");
+}
+
+function isGptImage2Model(model: string): boolean {
+  return model.includes("gpt-image-2");
+}
+
+function roundToMultiple(value: number, multiple: number): number {
+  return Math.max(multiple, Math.round(value / multiple) * multiple);
+}
+
+function buildGptImage2SizeFromAspectRatio(
+  ar: string | null,
+  quality: CliArgs["quality"],
+): string {
+  const parsed = ar ? parseAspectRatio(ar) : null;
+  const ratio = parsed ? parsed.width / parsed.height : 1;
+
+  if (!parsed || Math.abs(ratio - 1) < 0.1) {
+    const edge = quality === "2k" ? 2048 : 1024;
+    return `${edge}x${edge}`;
+  }
+
+  const targetLongEdge = quality === "2k" ? 2048 : 1024;
+  let width: number;
+  let height: number;
+
+  if (ratio > 1) {
+    width = targetLongEdge;
+    height = roundToMultiple(width / ratio, 16);
+  } else {
+    height = targetLongEdge;
+    width = roundToMultiple(height * ratio, 16);
+  }
+
+  while (width * height < 655_360) {
+    if (ratio > 1) {
+      width += 16;
+      height = roundToMultiple(width / ratio, 16);
+    } else {
+      height += 16;
+      width = roundToMultiple(height * ratio, 16);
+    }
+  }
+
+  return `${width}x${height}`;
+}
+
 export function getOpenAISize(
  model: string,
  ar: string | null,
@@ -37,6 +86,10 @@ export function getOpenAISize(
    return "1024x1024";
  }

+  if (isGptImage2Model(model)) {
+    return buildGptImage2SizeFromAspectRatio(ar, quality);
+  }
+
  const sizes: SizeMapping = isDalle3
    ? {
        square: "1024x1024",
@@ -127,6 +180,18 @@ export function getOpenAIResolution(
  return args.quality === "normal" ? "1K" : "2K";
 }

+function getOpenAIQuality(model: string, quality: CliArgs["quality"]): "standard" | "hd" | "medium" | "high" | null {
+  if (model.includes("dall-e-3")) {
+    return quality === "2k" ? "hd" : "standard";
+  }
+
+  if (isGptImageModel(model)) {
+    return quality === "2k" ? "high" : "medium";
+  }
+
+  return null;
+}
+
 export function getOrientationFromAspectRatio(ar: string): "landscape" | "portrait" | null {
  const parsed = parseAspectRatio(ar);
  if (!parsed) return null;
@@ -163,13 +228,53 @@ export function buildOpenAIGenerationsBody(
    size: args.size || getOpenAISize(model, args.aspectRatio, args.quality),
  };

-  if (model.includes("dall-e-3")) {
-    body.quality = args.quality === "2k" ? "hd" : "standard";
+  const quality = getOpenAIQuality(model, args.quality);
+  if (quality) {
+    body.quality = quality;
  }

  return body;
 }

+export function validateArgs(model: string, args: CliArgs): void {
+  if (!isGptImage2Model(model)) return;
+
+  if (args.aspectRatio && !args.size) {
+    const parsed = parseAspectRatio(args.aspectRatio);
+    if (!parsed) {
+      throw new Error(`Invalid gpt-image-2 aspect ratio: ${args.aspectRatio}`);
+    }
+    const ratio = parsed.width / parsed.height;
+    if (Math.max(ratio, 1 / ratio) > 3) {
+      throw new Error("gpt-image-2 aspect ratio must not exceed 3:1.");
+    }
+  }
+
+  if (!args.size) return;
+
+  const parsedSize = parsePixelSize(args.size);
+  if (!parsedSize) {
+    throw new Error(`Invalid gpt-image-2 --size: ${args.size}. Expected <width>x<height>.`);
+  }
+
+  const { width, height } = parsedSize;
+  const totalPixels = width * height;
+  const ratio = Math.max(width, height) / Math.min(width, height);
+
+  if (Math.max(width, height) > 3840) {
+    throw new Error("gpt-image-2 --size maximum edge length must be 3840px or less.");
+  }
+  if (width % 16 !== 0 || height % 16 !== 0) {
+    throw new Error("gpt-image-2 --size width and height must both be multiples of 16px.");
+  }
+  if (ratio > 3) {
+    throw new Error("gpt-image-2 --size long edge to short edge ratio must not exceed 3:1.");
+  }
+  if (totalPixels < 655_360 || totalPixels > 8_294_400) {
+    throw new Error("gpt-image-2 --size total pixels must be between 655,360 and 8,294,400.");
+  }
+}
+
 export async function generateImage(
  prompt: string,
  model: string,
@@ -198,7 +303,7 @@ export async function generateImage(
    }
    if (model.includes("dall-e-2") || model.includes("dall-e-3")) {
      throw new Error(
-        "Reference images with OpenAI in this skill require GPT Image models. Use --model gpt-image-1.5 (or another gpt-image model)."
+        "Reference images with OpenAI in this skill require GPT Image models. Use --model gpt-image-2 (or another gpt-image model)."
      );
    }
    const size = args.size || getOpenAISize(model, args.aspectRatio, args.quality);
@@ -283,8 +388,9 @@ async function generateWithOpenAIEdits(
  form.append("prompt", prompt);
  form.append("size", size);

-  if (model.includes("gpt-image")) {
-    form.append("quality", quality === "2k" ? "high" : "medium");
+  const openAIQuality = getOpenAIQuality(model, quality);
+  if (openAIQuality && openAIQuality !== "standard" && openAIQuality !== "hd") {
+    form.append("quality", openAIQuality);
  }

  for (const refPath of referenceImages) {
@@ -1,7 +1,7 @@
 ---
 name: baoyu-infographic
 description: Generate professional infographics with 21 layout types and 21 visual styles. Analyzes content, recommends layout×style combinations, and generates publication-ready infographics. Use when user asks to create "infographic", "信息图", "visual summary", "可视化", or "高密度信息大图".
-version: 1.56.1
+version: 1.57.1
 metadata:
  openclaw:
    homepage: https://github.com/JimLiu/baoyu-skills#baoyu-infographic
@@ -23,11 +23,17 @@ Concrete `AskUserQuestion` references below are examples — substitute the loca

 ## Image Generation Tools

-When this skill needs to render an image:
+When this skill needs to render an image, resolve the backend in this order:

- **Use whatever image-generation tool or skill is available** in the current runtime — e.g., Codex `imagegen`, Hermes `image_generate`, `baoyu-imagine`, or any equivalent the user has installed.
- **If multiple are available**, ask the user **once** at the start which to use (batch with any other initial questions).
- **If none are available**, tell the user and ask how to proceed.
+1. **Current-request override** — if the user names a specific backend in the current message, use it.
+2. **Saved preference** — if `EXTEND.md` sets `preferred_image_backend` to a backend available right now, use it.
+3. **Auto-select** (when the preference is `auto`, unset, or the pinned backend isn't available):
+   - If the current runtime exposes a native image tool (e.g., Codex `imagegen`, Hermes `image_generate`), use it. Runtime-native tools are preferred by default — agents that know their own tool inventory should surface the native one here.
+   - Otherwise, if exactly one non-native backend is installed (e.g., `baoyu-imagine`), use it.
+   - Otherwise (multiple non-native backends with no runtime-native tool), ask the user once — batch with any other initial questions.
+4. **If none are available**, tell the user and ask how to proceed.
+
+Setting `preferred_image_backend: ask` forces the step-3 prompt every run regardless of available backends. Users change the pinned backend via the `## Changing Preferences` section below.

 **Prompt file requirement (hard)**: write each image's full, final prompt to a standalone file under `prompts/` (naming: `NN-{type}-[slug].md`) BEFORE invoking any backend. The backend receives the prompt file (or its content); the file is the reproducibility record and lets you switch backends without regenerating prompts.

@@ -64,6 +70,15 @@ references:
 - If `usage: direct` AND the chosen backend accepts reference images (e.g., `baoyu-imagine` via `--ref`) → pass the file via the backend's ref parameter
 - Otherwise → embed extracted `style`/`palette` traits in the prompt text

+## Confirmation Policy
+
+Default behavior: **confirm before generation**.
+
+- Treat explicit skill invocation, a file path, a matched keyword shortcut, `EXTEND.md` defaults, and the documented default combination as **recommendation inputs only**. None of them authorizes skipping confirmation.
+- Do **not** start Step 5 or Step 6 until the user confirms the combination/aspect/language/backend choices.
+- Skip confirmation only when the current request explicitly says to do so, for example: `--no-confirm`, "直接生成", "不用确认", "跳过确认", "按默认出图", or equivalent wording.
+- If confirmation is skipped explicitly, state the assumed combination/aspect/language/backend in the next user-facing update before generating.
+
 ## Options

 | Option | Values |
@@ -72,6 +87,7 @@ references:
 | `--style` | 21 options (see Style Gallery), default: craft-handmade |
 | `--aspect` | Named: landscape (16:9), portrait (9:16), square (1:1). Custom: any W:H ratio (e.g., 3:4, 4:3, 2.35:1) |
 | `--lang` | en, zh, ja, etc. |
+| `--no-confirm` | Skip Step 4 only when the user explicitly requests direct generation without confirmation |
 | `--ref <files...>` | Reference images (file paths) for style / palette / composition / subject guidance |

 ## Layout Gallery (21)
@@ -152,11 +168,11 @@ Full definitions live at `references/styles/<style>.md`.
 | Educational Diagram | `hub-spoke` + `hand-drawn-edu` |
 | Process Tutorial | `linear-progression` + `hand-drawn-edu` |

-Default combination: `bento-grid` + `craft-handmade`.
+Default combination: `bento-grid` + `craft-handmade` (fallback recommendation only — per the [Confirmation Policy](#confirmation-policy), defaults never bypass Step 4).

 ## Keyword Shortcuts

-When the user's input contains these keywords, auto-select the layout and promote the listed styles to the top of Step 3 recommendations. Skip content-based layout inference for matched keywords. Append any `Prompt Notes` to the Step 5 prompt.
+When the user's input contains these keywords, use the mapped layout as the leading Step 3 recommendation and promote the listed styles to the top of the Step 3 list. Skip content-based layout inference for matched keywords. Append any `Prompt Notes` to the Step 5 prompt.

 | User Keyword | Layout | Recommended Styles | Default Aspect | Prompt Notes |
 |--------------|--------|--------------------|----------------|--------------|
@@ -201,7 +217,7 @@ Check EXTEND.md in priority order — the first one found wins:
 | Found | Read, parse, display a one-line summary |
 | Not found | Ask the user with `AskUserQuestion` (see `references/config/first-time-setup.md`) |

-**EXTEND.md supports**: preferred layout/style, default aspect ratio, custom style definitions, language preference.
+**EXTEND.md supports**: preferred layout/style, default aspect ratio, language preference, preferred image backend, custom style definitions.

 Schema: `references/config/preferences-schema.md`

@@ -231,7 +247,7 @@ See `references/structured-content-template.md` for detailed format.

 ### Step 3: Recommend Combinations

-**3.1 Check Keyword Shortcuts first**: If user input matches a keyword from the **Keyword Shortcuts** table, auto-select the associated layout and prioritize associated styles as top recommendations. Skip content-based layout inference.
+**3.1 Check Keyword Shortcuts first**: If user input matches a keyword from the **Keyword Shortcuts** table, use the associated layout as the leading recommendation and prioritize associated styles as top recommendations. Skip content-based layout inference.

 **3.2 Otherwise**, recommend 3-5 layout×style combinations based on:
 - Data structure → matching layout
@@ -241,6 +257,8 @@ See `references/structured-content-template.md` for detailed format.

 ### Step 4: Confirm Options

+**Hard gate**: this step is mandatory per the [Confirmation Policy](#confirmation-policy) — Steps 5–6 cannot start until the user confirms here (or explicitly opts out with `--no-confirm` / equivalent in the current request).
+
 Ask the user to confirm the questions below following the [User Input Tools](#user-input-tools) rule at the top of this file (batch into one call if the runtime supports multiple questions; otherwise ask one at a time in priority order).

 | Priority | Question | When | Options |
@@ -248,6 +266,7 @@ Ask the user to confirm the questions below following the [User Input Tools](#us
 | 1 | **Combination** | Always | 3+ layout×style combos with rationale |
 | 2 | **Aspect** | Always | Named presets (landscape/portrait/square) or custom W:H ratio (e.g., 3:4, 4:3, 2.35:1) |
 | 3 | **Language** | Only if source ≠ user language | Language for text content |
+| 4 | **Image Backend** | Only if step 3 of the `## Image Generation Tools` rule needs to ask (no runtime-native tool AND multiple non-native backends, OR `preferred_image_backend: ask`) | Available backends |

 ### Step 5: Generate Prompt → `prompts/infographic.md`

@@ -266,7 +285,7 @@ Combine:

 ### Step 6: Generate Image

-1. Select the backend via the `## Image Generation Tools` rule at the top: use whatever is available; if multiple, ask the user once. Do this once per session.
+1. Resolve the backend per the `## Image Generation Tools` rule at the top of this file.
 2. Ensure the full final prompt is persisted at `prompts/infographic.md` (already written in Step 5) BEFORE invoking the backend — the file is the reproducibility record.
 3. **Check for existing file**: Before generating, check if `infographic.png` exists
   - If exists: Rename to `infographic-backup-YYYYMMDD-HHMMSS.png`
@@ -275,7 +294,7 @@ Combine:

 ### Step 7: Output Summary

-Report: topic, layout, style, aspect, language, output path, files created.
+Report: topic, layout, style, aspect, language, image backend, output path, files created.

 ## References

@@ -285,6 +304,15 @@ Report: topic, layout, style, aspect, language, output path, files created.
 - `references/layouts/<layout>.md` - 21 layout definitions
 - `references/styles/<style>.md` - 21 style definitions

-## Extension Support
+## Changing Preferences

-Custom configurations via EXTEND.md. See **Step 1.1** for paths and supported options.
+EXTEND.md lives at the first matching path in Step 1.1. Three ways to change it:
+
+- **Edit directly** — open EXTEND.md and change fields. Full schema: `references/config/preferences-schema.md`.
+- **Reconfigure interactively** — delete EXTEND.md (or ask "reconfigure baoyu-infographic preferences" / "重新配置"). The next run re-triggers first-time setup.
+- **Common one-line edits**:
+  - `preferred_image_backend: auto` — default; runtime-native tool wins, falls back to the only installed backend, asks only if multiple non-native are present.
+  - `preferred_image_backend: codex-imagegen` — pin to Codex's built-in.
+  - `preferred_image_backend: baoyu-imagine` — pin to the baoyu-imagine skill.
+  - `preferred_image_backend: ask` — confirm backend every run.
+  - `preferred_layout: dense-modules`, `preferred_style: morandi-journal`, `preferred_aspect: portrait`, `language: zh` — shift the Step-3 recommendations and Step-4 defaults (per [Confirmation Policy](#confirmation-policy), these never bypass Step 4).
@@ -7,7 +7,7 @@ description: First-time setup flow for baoyu-infographic preferences

 ## Overview

-When no EXTEND.md is found, guide the user through preference setup before generating any infographic.
+When no EXTEND.md is found, guide the user through preference setup before generating any infographic. Saved preferences shift Step-3 recommendations and Step-4 defaults only — they never bypass Step 4 confirmation (see the `## Confirmation Policy` section in SKILL.md).

 **⛔ BLOCKING OPERATION**: This setup MUST complete before ANY other workflow steps. Do NOT:
 - Ask about source content or topic
@@ -141,13 +141,13 @@ preferred_layout: [selected layout or null]
 preferred_style: [selected style or null]
 preferred_aspect: [landscape|portrait|square|null]
 language: [selected language or null]
+preferred_image_backend: auto
 custom_styles: []
 ---
 ```

+`preferred_image_backend: auto` is the baked-in default — first-time setup never asks about it. The `## Image Generation Tools` rule in SKILL.md then picks the runtime-native tool (Codex `imagegen`, Hermes `image_generate`, etc.) when one is available, and falls back to installed backends like `baoyu-imagine`.
+
 ## Modifying Preferences Later

-Users can edit EXTEND.md directly or trigger setup again:
- Delete EXTEND.md to re-trigger setup
- Edit YAML frontmatter for quick changes
- Full schema: `references/config/preferences-schema.md`
+See the `## Changing Preferences` section in `SKILL.md` for the canonical list of common edits (pin backend, change layout/style defaults, retrigger setup). Full schema: `references/config/preferences-schema.md`.
@@ -17,6 +17,8 @@ preferred_aspect: null    # landscape|portrait|square|null  (custom W:H also acc

 language: null            # zh|en|ja|ko|null (null = auto-detect from source)

+preferred_image_backend: auto   # auto|ask|<backend-id>
+
 custom_styles:            # extra style definitions merged with the 21 built-ins
  - name: my-brand
    description: "Short description shown in Step 3 recommendations"
@@ -33,8 +35,21 @@ custom_styles:            # extra style definitions merged with the 21 built-ins
 | `preferred_style` | string\|null | null | Pre-selected style — surfaces as the top recommendation in Step 3 |
 | `preferred_aspect` | string\|null | null | Default aspect for Step 4 (named preset or W:H string) |
 | `language` | string\|null | null | Output language (null = auto-detect from source content) |
+| `preferred_image_backend` | string | `auto` | Image backend selection. `auto` = prefer runtime-native tool, fall back to the only installed backend, ask if multiple non-native are present. `ask` = always confirm on every run. `<backend-id>` (e.g., `codex-imagegen`, `baoyu-imagine`, `image_generate`) = pin this backend when available; fall back to `auto` when it isn't. Absent = `auto`. |
 | `custom_styles` | array | [] | Additional styles available alongside the 21 built-ins |

+Backend resolution logic is documented in the `## Image Generation Tools` section of `SKILL.md`. This doc only defines the field.
+
+All fields in this schema are defaults only — they shape Step-3 recommendations and Step-4 defaults but never bypass Step 4 confirmation (see the `## Confirmation Policy` section in SKILL.md).
+
+Example backend ids:
+
+| Value | Meaning |
+|-------|---------|
+| `codex-imagegen` | Codex built-in `imagegen` tool |
+| `baoyu-imagine` | `baoyu-imagine` skill / script backend |
+| `image_generate` | Generic runtime image tool such as Hermes |
+
 ## Layout Options

 See the **Layout Gallery (21)** table in `SKILL.md` for the canonical list. Common picks:
@@ -87,6 +102,8 @@ language: zh
 ---
 ```

+`preferred_image_backend` is omitted above; absence is treated as `auto`.
+
 ## Example: Full Preferences

 ```yaml
@@ -99,6 +116,8 @@ preferred_aspect: portrait

 language: zh

+preferred_image_backend: codex-imagegen
+
 custom_styles:
  - name: my-brand
    description: "Brand-aligned warm pastel infographic"
@@ -1,7 +1,7 @@
 ---
 name: baoyu-post-to-x
 description: Posts content and articles to X (Twitter). Supports regular posts with images/videos and X Articles (long-form Markdown). Uses real Chrome with CDP to bypass anti-automation. Use when user asks to "post to X", "tweet", "publish to Twitter", or "share on X".
-version: 1.56.1
+version: 1.56.2
 metadata:
  openclaw:
    homepage: https://github.com/JimLiu/baoyu-skills#baoyu-post-to-x
@@ -5,6 +5,7 @@ import os from 'node:os';
 import path from 'node:path';
 import process from 'node:process';
 import { createHash } from 'node:crypto';
+import { pathToFileURL } from 'node:url';

 import frontMatter from 'front-matter';
 import hljs from 'highlight.js/lib/common';
@@ -458,7 +459,9 @@ async function main(): Promise<void> {
  }
 }

-await main().catch((err) => {
-  console.error(`Error: ${err instanceof Error ? err.message : String(err)}`);
-  process.exit(1);
-});
+if (process.argv[1] && import.meta.url === pathToFileURL(process.argv[1]).href) {
+  await main().catch((err) => {
+    console.error(`Error: ${err instanceof Error ? err.message : String(err)}`);
+    process.exit(1);
+  });
+}
@@ -1,7 +1,7 @@
 ---
 name: baoyu-slide-deck
 description: Generates professional slide deck images from content. Creates outlines with style instructions, then generates individual slide images. Use when user asks to "create slides", "make a presentation", "generate deck", "slide deck", or "PPT".
-version: 1.56.1
+version: 1.56.2
 metadata:
  openclaw:
    homepage: https://github.com/JimLiu/baoyu-skills#baoyu-slide-deck
@@ -27,14 +27,31 @@ Concrete `AskUserQuestion` references below are examples — substitute the loca

 ## Image Generation Tools

-When this skill needs to render an image:
+When this skill needs to render an image, resolve the backend in this order:

- **Use whatever image-generation tool or skill is available** in the current runtime — e.g., Codex `imagegen`, Hermes `image_generate`, `baoyu-imagine`, or any equivalent the user has installed.
- **If multiple are available**, ask the user **once** at the start which to use (batch with any other initial questions).
- **If none are available**, tell the user and ask how to proceed.
+1. **Current-request override** — if the user names a specific backend in the current message, use it.
+2. **Saved preference** — if `EXTEND.md` sets `preferred_image_backend` to a backend available right now, use it.
+3. **Auto-select** (when the preference is `auto`, unset, or the pinned backend isn't available):
+   - If the current runtime exposes a native image tool (e.g., Codex `imagegen`, Hermes `image_generate`), use it. Runtime-native tools are preferred by default — agents that know their own tool inventory should surface the native one here.
+   - Otherwise, if exactly one non-native backend is installed (e.g., `baoyu-imagine`), use it.
+   - Otherwise (multiple non-native backends with no runtime-native tool), ask the user once — batch with any other initial questions.
+4. **If none are available**, tell the user and ask how to proceed.
+
+Setting `preferred_image_backend: ask` forces the step-3 prompt every run regardless of available backends. Users change the pinned backend via the `## Changing Preferences` section below.

 **Prompt file requirement (hard)**: write each image's full, final prompt to a standalone file under `prompts/` (naming: `NN-slide-[slug].md`) BEFORE invoking any backend. The file is the reproducibility record and lets you switch backends without regenerating prompts.

+Concrete tool names (`imagegen`, `image_generate`, `baoyu-imagine`) above are examples — substitute the local equivalents under the same rule.
+
+## Confirmation Policy
+
+Default behavior: **confirm before generation**.
+
+- Treat explicit skill invocation, a file path, matched signals/presets, and `EXTEND.md` defaults as **recommendation inputs only**. None of them authorizes skipping confirmation.
+- Do **not** start Step 3 or later until the user completes Step 2.
+- Skip confirmation only when the current request explicitly says to do so, for example: "直接生成", "不用确认", "跳过确认", "按默认出幻灯片", or equivalent wording.
+- If confirmation is skipped explicitly, state the assumed style / audience / slide-count / language / backend in the next user-facing update before generating.
+
 ## Language

 Respond in the user's language across questions, progress reports, error messages, and the completion summary. Keep technical tokens (style names, file paths, code) in English.
@@ -213,6 +230,8 @@ Save findings to `analysis.md`: topic, audience, signals, recommended style and

 ### Step 2: Confirmation ⚠️ REQUIRED

+**Hard gate**: this step is mandatory per the [Confirmation Policy](#confirmation-policy) — Steps 3+ cannot start until the user confirms here (or explicitly opts out with "直接生成" / equivalent wording in the current request).
+
 **Round 1 (always)** — batch five questions in one `AskUserQuestion` call: style, audience, slide count, review-outline?, review-prompts?. Verbatim options in `references/confirmation.md`.

 Summary displayed before the questions:
@@ -320,4 +339,14 @@ See `references/modification-guide.md` for full details.
 - For sensitive public figures, prefer stylized alternatives to avoid likeness issues.
 - Maintain visual consistency via the session ID when the backend supports it.

-Custom configurations via EXTEND.md. See Step 1.1 for paths and schema.
+## Changing Preferences
+
+EXTEND.md lives at the first matching path listed in Step 1.1. Two ways to change it:
+
+- **Edit directly** — open EXTEND.md and change fields. Full schema: `references/config/preferences-schema.md`.
+- **Common one-line edits**:
+  - `preferred_image_backend: auto` — default; runtime-native tool wins, falls back to the only installed backend, asks only if multiple non-native are present.
+  - `preferred_image_backend: codex-imagegen` — pin to Codex's built-in.
+  - `preferred_image_backend: baoyu-imagine` — pin to the baoyu-imagine skill.
+  - `preferred_image_backend: ask` — confirm backend every run.
+  - `preferred_style: blueprint`, `preferred_audience: experts`, `language: zh`.
@@ -12,6 +12,7 @@ style: blueprint              # Preset name OR "custom"
 audience: general             # beginners | intermediate | experts | executives | general
 language: auto                # auto | en | zh | ja | etc.
 review: true                  # true = review outline before generation
+preferred_image_backend: auto # auto | ask | <backend-id>

 ## Custom Dimensions (only when style: custom)
 dimensions:
@@ -40,6 +41,7 @@ custom_styles:
 | `audience` | string | `general` | Default target audience |
 | `language` | string | `auto` | Output language (auto = detect from input) |
 | `review` | boolean | `true` | Show outline review before generation |
+| `preferred_image_backend` | string | `auto` | Image backend selection. `auto` = prefer runtime-native tool, fall back to the only installed backend, ask if multiple non-native are present. `ask` = always confirm on every run. `<backend-id>` (e.g., `codex-imagegen`, `baoyu-imagine`, `image_generate`) = pin this backend when available; fall back to `auto` when it isn't. Absent = `auto`. Resolution logic is documented in `SKILL.md`'s `## Image Generation Tools` section. |

 ### Custom Dimensions

@@ -1,7 +1,7 @@
 ---
 name: baoyu-xhs-images
 description: "[Deprecated: use baoyu-image-cards] Generates Xiaohongshu (Little Red Book) image card series with 12 visual styles, 8 layouts, and 3 color palettes. Breaks content into 1-10 cartoon-style image cards optimized for XHS engagement. Use when user mentions \"小红书图片\", \"XHS images\", \"RedNote infographics\", \"小红书种草\", \"小绿书\", \"微信图文\", \"微信贴图\", or wants social media infographic series for Chinese platforms."
-version: 1.56.1
+version: 1.56.2
 metadata:
  openclaw:
    homepage: https://github.com/JimLiu/baoyu-skills#baoyu-xhs-images
@@ -24,14 +24,31 @@ Concrete `AskUserQuestion` references below are examples — substitute the loca

 ## Image Generation Tools

-When this skill needs to render an image:
+When this skill needs to render an image, resolve the backend in this order:

- **Use whatever image-generation tool or skill is available** in the current runtime — e.g., Codex `imagegen`, Hermes `image_generate`, `baoyu-imagine`, or any equivalent the user has installed.
- **If multiple are available**, ask the user **once** at the start which to use (batch with any other initial questions).
- **If none are available**, tell the user and ask how to proceed.
+1. **Current-request override** — if the user names a specific backend in the current message, use it.
+2. **Saved preference** — if `EXTEND.md` sets `preferred_image_backend` to a backend available right now, use it.
+3. **Auto-select** (when the preference is `auto`, unset, or the pinned backend isn't available):
+   - If the current runtime exposes a native image tool (e.g., Codex `imagegen`, Hermes `image_generate`), use it. Runtime-native tools are preferred by default — agents that know their own tool inventory should surface the native one here.
+   - Otherwise, if exactly one non-native backend is installed (e.g., `baoyu-imagine`), use it.
+   - Otherwise (multiple non-native backends with no runtime-native tool), ask the user once — batch with any other initial questions.
+4. **If none are available**, tell the user and ask how to proceed.
+
+Setting `preferred_image_backend: ask` forces the step-3 prompt every run regardless of available backends. Users change the pinned backend via the `## Changing Preferences` section below.

 **Prompt file requirement (hard)**: write each image's full, final prompt to a standalone file under `prompts/` (naming: `NN-{type}-[slug].md`) BEFORE invoking any backend. The file is the reproducibility record and lets you switch backends without regenerating prompts.

+Concrete tool names (`imagegen`, `image_generate`, `baoyu-imagine`) above are examples — substitute the local equivalents under the same rule.
+
+## Confirmation Policy
+
+Default behavior: **confirm before generation**.
+
+- Treat explicit skill invocation, a file path, matched signals/presets, and `EXTEND.md` defaults as **recommendation inputs only**. None of them authorizes skipping confirmation.
+- Do **not** start Step 3 until the user completes Step 2.
+- Skip confirmation only when the current request explicitly says to do so, for example: `--yes`, "直接生成", "不用确认", "跳过确认", "按默认出图", or equivalent wording.
+- If confirmation is skipped explicitly, state the assumed strategy / style / layout / palette / count / backend in the next user-facing update before generating.
+
 ## Language

 Respond in the user's language across questions, progress, errors, and completion summary. Keep technical tokens (style names, file paths, code) in English.
@@ -292,6 +309,8 @@ Check these paths in order; first hit wins:

 ### Step 2: Smart Confirm ⚠️ REQUIRED

+**Hard gate**: this step is mandatory per the [Confirmation Policy](#confirmation-policy) — Step 3 cannot start until the user confirms here (or explicitly opts out with `--yes` / equivalent wording in the current request).
+
 Goal: present the auto-recommended plan and let the user confirm or adjust. Skip this step entirely under `--yes` — proceed with Path A using the analysis and any CLI overrides.

 **Display summary** before asking:
@@ -418,4 +437,16 @@ Always update the prompt file before regenerating — it's the source of truth a
 - For sensitive public figures, use stylized cartoon alternatives.
 - Smart Confirm (Step 2) is required; Detailed mode adds a second confirmation (2a + 2c).

-Custom configurations via EXTEND.md. See Step 0 for paths and schema.
+## Changing Preferences
+
+EXTEND.md lives at the first matching path listed in Step 0. Three ways to change it:
+
+- **Edit directly** — open EXTEND.md and change fields. Full schema: `references/config/preferences-schema.md`.
+- **Reconfigure interactively** — delete EXTEND.md (or ask "reconfigure baoyu-xhs-images preferences" / "重新配置"). The next run re-triggers first-time setup.
+- **Common one-line edits**:
+  - `preferred_image_backend: auto` — default; runtime-native tool wins, falls back to the only installed backend, asks only if multiple non-native are present.
+  - `preferred_image_backend: codex-imagegen` — pin to Codex's built-in.
+  - `preferred_image_backend: baoyu-imagine` — pin to the baoyu-imagine skill.
+  - `preferred_image_backend: ask` — confirm backend every run.
+  - `preferred_style: notion`, `preferred_layout: dense`, `preferred_palette: macaron`, `language: zh`.
+  - `watermark.enabled: true` + `watermark.content: "@handle"` — add a watermark.
@@ -110,13 +110,13 @@ preferred_style:
  description: ""
 preferred_layout: null
 language: null
+preferred_image_backend: auto
 custom_styles: []
 ---
 ```

+`preferred_image_backend: auto` is the baked-in default — first-time setup does not ask about it. The `## Image Generation Tools` rule in SKILL.md then picks the runtime-native tool (Codex `imagegen`, Hermes `image_generate`, etc.) when available, and falls back to installed backends.
+
 ## Modifying Preferences Later

-Users can edit EXTEND.md directly or run setup again:
- Delete EXTEND.md to trigger setup
- Edit YAML frontmatter for quick changes
- Full schema: `config/preferences-schema.md`
+See the `## Changing Preferences` section in `SKILL.md` for the canonical list of common edits (pin backend, change defaults, retrigger setup). Full schema: `preferences-schema.md`.
@@ -24,6 +24,8 @@ preferred_layout: null    # sparse|balanced|dense|list|comparison|flow

 language: null            # zh|en|ja|ko|auto

+preferred_image_backend: auto  # auto|ask|<backend-id>
+
 custom_styles:
  - name: my-style
    description: "Style description"
@@ -49,6 +51,7 @@ custom_styles:
 | `preferred_style.description` | string | "" | Custom notes/override |
 | `preferred_layout` | string | null | Layout preference or null |
 | `language` | string | null | Output language (null = auto-detect) |
+| `preferred_image_backend` | string | `auto` | Image backend selection. `auto` = prefer runtime-native tool, fall back to the only installed backend, ask if multiple non-native are present. `ask` = always confirm on every run. `<backend-id>` (e.g., `codex-imagegen`, `baoyu-imagine`, `image_generate`) = pin this backend when available; fall back to `auto` when it isn't. Absent = `auto`. Resolution logic is documented in `SKILL.md`'s `## Image Generation Tools` section. |
 | `custom_styles` | array | [] | User-defined styles |

 ## Position Options
@@ -104,6 +107,8 @@ preferred_layout: dense

 language: zh

+preferred_image_backend: codex-imagegen
+
 custom_styles:
  - name: corporate
    description: "Professional B2B style"
Author	SHA1	Message	Date
Jim Liu 宝玉	505a7e10ce	chore: release v1.113.0	2026-04-25 15:03:22 -05:00
Jim Liu 宝玉	6d063734ae	feat(baoyu-imagine): add DashScope Wan 2.7 image model support (#141 ) * feat(baoyu-imagine): add DashScope Wan 2.7 image model support Closes #139. Adds the new `wan2.7-image-pro` and `wan2.7-image` model family to the DashScope provider so users can call Wan 2.7 directly through the official Aliyun (Bailian) API instead of going through Replicate. - Register `wan2.7-image-pro` and `wan2.7-image` as a new `wan27` family in the DashScope provider with their own size resolution rules: pixel range `[768768, 40964096]` for `wan2.7-image-pro` text-to-image, `[768768, 20482048]` for `wan2.7-image-pro` with refs and for the base `wan2.7-image` model in any mode, with aspect ratios validated against the documented `[1:8, 8:1]` band. - Allow up to 9 reference images per request (image editing / multi-image fusion). Local files are inlined as base64 data URLs; `http(s)://` paths are forwarded as-is. Other DashScope models still reject `--ref` with a hint to switch to a wan2.7 model or another provider. - Drop `prompt_extend` from the request body for the Wan 2.7 family (not part of the Wan 2.7 API surface) and skip the Qwen-only negative prompt for this family. - Allow `--provider dashscope --ref ...` in `detectProvider` so users can opt into Wan 2.7 reference workflows, while keeping Wan 2.7 out of the auto-detect ref priority list. - Add provider, reference, and usage-example documentation, plus unit tests covering family routing, size derivation across the three pixel-budget modes, ratio rejection, explicit-size validation, and the new `--provider dashscope` ref opt-in path. Made-with: Cursor * fix(baoyu-imagine): force n=1 for DashScope wan2.7 to avoid silent multi-image billing Cross-checked the implementation against the official Wan 2.7 image generation & editing API reference and found that the API defaults `parameters.n` to 4 in non-collage mode (1-4 range, billed per image). baoyu-imagine has single-image save semantics — only the first image in the response is kept — so without an explicit `n: 1` users would silently pay for 3 discarded images per request. - Always send `parameters.n: 1` in the wan2.7 request body - Reject `--n > 1` for wan2.7 with a clear error pointing at the single-image save semantics - Add tests asserting the request body shape (n=1, no prompt_extend, no negative_prompt) and the --n>1 rejection - Document the defaults-vs-skill mismatch in the dashscope reference Made-with: Cursor * Fix DashScope Wan 2.7 review feedback	2026-04-25 14:54:08 -05:00
Jim Liu 宝玉	31d728b505	chore: release v1.112.0	2026-04-24 02:15:57 -05:00
Jim Liu 宝玉	f6d5df0594	fix(baoyu-post-to-x): add entry point guard to md-to-html.ts for module import compatibility Wrap main() in an import.meta.url check so that importing parseMarkdown from x-article.ts no longer triggers the CLI entry point. Mirrors the same fix applied to baoyu-post-to-weibo.	2026-04-24 02:15:30 -05:00
Jim Liu 宝玉	4bd5fe573e	feat(baoyu-article-illustrator): default to sketch-notes educational infographic style Make `hand-drawn-edu` (infographic + sketch-notes + macaron) the universal fallback preset when content analysis surfaces no strong signal. Rework sketch-notes style spec around warm cream paper + black hand-drawn lines + soft pastel section blocks. - Change `hand-drawn-edu` preset type from flowchart to infographic; add `hand-drawn-edu-flow` (flowchart) and `hand-drawn-edu-compare` (comparison) as variants for users who need those layouts in the same warm style - Elevate `sketch-notes` to primary style across infographic / flowchart / comparison / framework auto-selection; add sketch-notes column to Type x Style compatibility matrix - Rewrite sketch-notes.md: macaron pastel palette, canonical single-page layout (title / sectioned boxes / takeaway), diagram-only rule - Add infographic + sketch-notes + macaron prompt block to prompt-construction.md - Update workflow, style-presets, and first-time-setup defaults to match	2026-04-24 02:15:25 -05:00
Jim Liu 宝玉	8c17d77209	chore: release v1.111.1	2026-04-21 16:59:35 -05:00
Jim Liu 宝玉	fb749aae36	docs(baoyu-article-illustrator): add Confirmation Policy State that defaults / signals / EXTEND.md preferences are recommendation inputs only — Step 3 is mandatory unless the current request opts out explicitly.	2026-04-21 16:59:11 -05:00
Jim Liu 宝玉	a23f2e6a4b	docs(baoyu-xhs-images): add Confirmation Policy Sync Confirmation Policy with baoyu-image-cards (deprecated skill kept functional per CLAUDE.md).	2026-04-21 16:59:11 -05:00
Jim Liu 宝玉	ae32c56a3c	docs(baoyu-image-cards): add Confirmation Policy State that defaults / signals / EXTEND.md preferences are recommendation inputs only — Step 2 is mandatory unless the current request opts out via `--yes` / equivalent wording.	2026-04-21 16:59:07 -05:00
Jim Liu 宝玉	4b2ad3ad18	docs(baoyu-slide-deck): add Confirmation Policy State that defaults / signals / EXTEND.md preferences are recommendation inputs only — Step 2 is mandatory unless the current request opts out explicitly.	2026-04-21 16:59:07 -05:00
Jim Liu 宝玉	512b6d2e15	docs(baoyu-cover-image): add Confirmation Policy State that defaults / keywords / EXTEND.md preferences are recommendation inputs only — Step 2 is mandatory unless the current request opts out via `--quick` / `quick_mode: true` / equivalent wording.	2026-04-21 16:59:03 -05:00
Jim Liu 宝玉	60d0b9c95b	docs(baoyu-infographic): add Confirmation Policy as single source of truth Consolidate the confirmation rule into one section so defaults, keywords, and EXTEND.md preferences are treated as recommendation inputs only — never as authorization to skip Step 4. Remove the repeated reminders that were scattered across Step 5, Step 6, Default combination, Keyword Shortcuts, and the preferences docs.	2026-04-21 16:58:58 -05:00
Jim Liu 宝玉	acf1d42a64	chore: release v1.111.0	2026-04-21 16:44:01 -05:00
Jim Liu 宝玉	adb783ae6d	refactor: unify image-backend resolution with preferred_image_backend preference Replaces the stateless ask-once backend rule with a 4-step resolution (current-request override > saved preference > auto-select > ask) and adds a single `preferred_image_backend` field (auto \| ask \| <backend-id>) to every image-consuming skill's EXTEND.md schema. Runtime-native tools (Codex `imagegen`, Hermes `image_generate`, ...) win by default; absent field equals `auto` so existing user EXTEND.md files stay valid with no schema version bump. Each image-consuming skill also gains a top-level `## Changing Preferences` section as a first-class surface for pinning a backend and editing common one-line preferences. Applies across: - baoyu-infographic (reverts in-progress two-field image_backend_mode design) - baoyu-comic - baoyu-cover-image - baoyu-image-cards - baoyu-article-illustrator - baoyu-slide-deck - baoyu-xhs-images - docs/image-generation-tools.md (canonical author-side doc)	2026-04-21 16:43:58 -05:00
Jim Liu 宝玉	a4bd37c390	chore: release v1.110.0	2026-04-21 15:15:19 -05:00