| name | pr-to-xhs |
| description | Generate Xiaohongshu (小红书) blog content and infographic images from a GitHub PR. Use when user provides a GitHub PR URL or asks to create XHS content from a PR. |
| version | 1.0.0 |
PR to Xiaohongshu Blog Generator
Generate Xiaohongshu blog content and infographic images from a GitHub PR.
Usage
/pr-to-xhs <github-pr-url>
/pr-to-xhs https://github.com/vllm-project/vllm-omni/pull/1689
Workflow
Step 1: Fetch PR Information
gh pr view <pr-number> --repo <owner/repo> --json title,body,author,additions,deletions,files,commits
Step 2: Create Output Directory
blogs/pr<pr-number>/xiaohongshu/images/xhs-images/<topic-slug>/
├── source-<slug>.md # Source content
├── analysis.md # Content analysis
├── outline.md # Final outline
├── post.json # XHS post configuration
├── prompts/ # Image prompts
│ ├── 01-cover-<slug>.md
│ ├── 02-content-<slug>.md
│ └── ...
├── 01-cover-<slug>.png # Generated images
├── 02-content-<slug>.png
└── ...
Step 3: Analyze Content
Create analysis.md with:
- Content type classification (干货/种草/测评/教程)
- Core selling points
- Target audience
- Recommended XHS style and layout
Step 4: Generate Outline
Create outline.md with:
- 4 images: Cover + 2 Content + Ending
- Style: notion (hand-drawn line art)
- Layout: balanced
Step 5: Generate Images
Use glm-image skill (Zhipu AI) to generate:
- P1 Cover: Hook title + visual impact
- P2 Core Features: 3 key points
- P3 Technical Details: Code examples, parameters
- P4 Ending: CTA + GitHub link
Step 6: Create post.json
{
"title": "<title>",
"content": "<body with hashtags>",
"images": ["01-cover.png", "02-content.png", ...],
"tags": ["tag1", "tag2", ...],
"source": {
"type": "github_pr",
"url": "<pr-url>",
"pr_number": <number>
}
}
Image Generation
Default configuration:
- Provider: Zhipu AI (OpenAI-compatible)
- Model: glm-4-0520
- Aspect Ratio: 3:4 (Xiaohongshu optimized)
- Quality: 2k
- Style: notion (minimalist hand-drawn line art)
Style Guidelines
notion style elements:
- Clean light background (#FAFAFA)
- Hand-drawn line art aesthetic
- No gradients or shadows
- Simple geometric shapes
- Professional yet approachable
Color palette:
- Background: #FAFAFA
- Main text: #1A1A1A
- Accent blue: #3B82F6
- Accent green: #10B981
- Line art: #E5E7EB
Example
Input:
/pr-to-xhs https://github.com/vllm-project/vllm-omni/pull/1689
Output:
blogs/pr1689/xiaohongshu/images/xhs-images/hunyuan-image3-npu-support/- 4 PNG images (3:4 aspect ratio)
- post.json with title, content, tags
