Guizang Social Card Skill – Open-Source Xiaohongshu & WeChat Image Card Skill
Executive Summary:
Guizang Social Card Skill is an open-source skill by independent developer Guizang (op7418) for Claude Code, Codex, and similar AI agent environments. It generates magazine-quality social image cards ...
1. What Is Guizang Social Card Skill
Guizang Social Card Skill is an open-source skill by independent developer Guizang (op7418) for Claude Code, Codex, and similar AI agent environments. It generates magazine-quality social image cards from single-file HTML + CSS rendered with Playwright—Xiaohongshu 3:4 verticals, WeChat Official Account 21:9 headers, and 1:1 share cards. Two visual systems (Editorial and Swiss), 28 layout skeletons, and 10 theme presets drive a structured seven-step workflow from content input to finished PNG—fixing the "AI look" and layout chaos common in social graphics.
Technical positioning and domain: Agent toolchain for content creation and visual design—encoding Monocle/Kinfolk/Cereal-style typography as agent-callable HTML/CSS templates instead of diffusion models or manual design apps. Structured text lets agents control grids, color, tracking, and image safe zones end-to-end.
Development background: Guizang is an active open-source contributor in AI apps and agent skills. Motivation: strong LLM copy, weak social visuals—manual design or crude templates lack consistent aesthetics. A programmable layout system lets agents ship on-brand cards with content.
Core value: (1) Reduces generic "AI slop" via restrained palettes, precise typography, and grids that stand out in feeds. (2) Category-aware routing—travel vs reviews vs food pick layouts and tones automatically. (3) Reliable text-on-image via subject avoidance, luminance sampling, and adaptive line breaks—no text over faces or unreadable contrast.
Technical characteristics: Plain HTML/CSS agents can edit—no frontend build chain; CSS Grid hierarchy without decorative shadows; image-overlay avoid zones; dynamic font sizing; SKILL.md orchestrates the full pipeline.
2. Key Features
Dual visual systems: Editorial (narrative, travel, lifestyle—full-bleed photos, airy type) vs Swiss (reviews, data, methodology—grids, device frames, spec comparisons)—switch by style name.
Three canvas sizes: Native Xiaohongshu 3:4, WeChat 21:9 header, 1:1 share—auto-selected per platform, no manual crop.
28 layout skeletons: Editorial M01–M16 (covers, pipelines, before/after, quotes); Swiss S01–S12 (KPI towers, bar charts, matrix + hero)—agent picks by content type.
10 theme presets + smart color: Editorial six themes (ink classic, indigo porcelain, forest ink, kraft, dune, midnight ink); Swiss four anchors (Klein blue, lemon yellow, lime, safety orange)—fixed palettes (no custom hex) for consistency; mood-based theme suggestions.
Three-step text-on-image: (1) image-overlay marks faces/products/text as avoid; (2) luminance sampling in safe zones picks B/W text and overlay depth; (3) dynamic font size and wrap—force deeper overlay if contrast <4.5.
Auto image sourcing: Unsplash → Pexels → Flickr CC → Wallhaven with auto SOURCES.md for licensing.
Maps and screenshot polish: MapLibre + OSM tiles with pins/routes; nine WebP texture backgrounds;
.frame-shot,.device-browser,.device-phonefor product shots.WebGL ink hero + 11 category rules: Dynamic ink backgrounds on Editorial heroes; auto routing for travel, career, film, reviews, books, food, games, makeup, fitness, home, fashion.
3. How to Use
Requirements: Node.js 18+, Playwright Chromium (
npx playwright install chromium). Linux/macOS preferred; agent needs shell access.Install skill:
npx skills add https://github.com/op7418/guizang-social-card-skill --skill guizang-social-card-skillOr manual:
git clone https://github.com/op7418/guizang-social-card-skill.git ~/.claude/skills/guizang-social-card-skillAuthor content: Create Markdown under
content/per SKILL.md, e.g.:--- platform: xiaohongshu style: editorial theme: ink-classic layout: M03 images: - https://example.com/photo.jpg --- # Title Body...Render:
node render.mjs—Playwright screenshots HTML templates tooutput/PNG with avoidance and theming applied.Validate:
validate-social-deck.mjschecks overflow, font caps, footer collisions, etc.Notes: Network needed first run for browser and stock images; use ≥1200px-wide custom photos; themes lock hex values; maps need online OSM tiles.
4. Pros and Cons
| Pros |
|---|
| Magazine-grade layout: Monocle/Kinfolk-inspired restraint beats emoji-heavy AI cards—real editorial hierarchy. |
| Category-aware routing: 11 vertical rules match layout/theme/stock style to content—not one template for all. |
| Reliable text-on-image: Avoid + luminance + wrap fixes face overlap and low contrast automatically. |
| Licensed stock pipeline: Auto fetch + SOURCES.md—no hunt for royalty-free assets. |
5. Comparison with Similar Tools
| Dimension | Guizang Social Card Skill | Brand Social Design |
|---|---|---|
| Author | Guizang (op7418) | Om Rajguru |
| Platforms | Xiaohongshu 3:4, WeChat 21:9, 1:1 | Instagram, OG, email, stories |
| Visual systems | Editorial + Swiss, 28 skeletons | Dot grid + hard shadow cards |
| Brand | 10 locked themes | Onboarding interview, memorized palette |
| Category rules | 11 vertical routers | None |
| Text-on-image | 3-step smart overlay | Fixed text regions |
| Stock images | Auto multi-source + attribution | User-supplied |
| Deploy | HTML + Playwright, agent-native | Node build (Vite/Webpack) |
Selection advice: Xiaohongshu/WeChat magazine look with category smarts—Guizang. Broader Instagram/OG brand memory—Brand Social Design (heavier setup). Simple Twitter/LinkedIn cards—SocialScreenshots (lighter, fewer features).
6. Editor's Review
Guizang isn't another image model—it's a programmable design system. Magazine rules as HTML/CSS give agents pixel control without diffusion text glitches—ideal for high-volume, consistent social matrix content.
It collapses "write post + design images" into one agent pass with text-on-image quality rivals lack. Category routing and stock automation help non-designers.
Audience: agent power users, social ops, reviewers, travel creators, WeChat editors; integrators chaining RAG → cards.
Future: more platforms/skeletons via community; offline LLM + local assets would widen adoption.
— −0.5 for agent dependency, CJK fine-tuning, rigid themes; still strongly recommended for CN social workflows.
7. Use Cases
Long article → Xiaohongshu carousel: Split analysis or book notes into 3–9 Editorial or Swiss slides in one agent run.
Product reviews: Swiss + Klein blue—device frames, KPI bars, spec matrices for gadgets/beauty/home.
Travel guides: Editorial full-bleed + MapLibre pins/routes from place lists.
WeChat covers: Simultaneous 21:9 header and 1:1 share from one content file.
Brand matrix: One copy → platform-sized, on-brand batches for Xiaohongshu + Official Account.
8. FAQ
Q: Instagram/Twitter support?
A: Native CN platforms only; extend by changing canvas CSS—community templates TBD.
Q: GPU required?
A: No—Playwright CPU screenshots; 4GB+ RAM recommended.
Q: Custom colors/fonts?
A: Themes lock hex for consistency; edit CSS variables at your own risk. System fonts default; web fonts need extra loading.
Q: Output resolution?
A: Xiaohongshu 1080×1440; WeChat header 1920×822; share 1080×1080—tweak viewport in render.mjs carefully.
Q: Uneven Chinese spacing?
A: Prefer short lines (~20 chars) or adjust word-spacing/letter-spacing; CJK engine improvements planned.
Q: Stock licensing?
A: Unsplash/Pexels/Flickr CC/Wallhaven—verify SOURCES.md and each license for commercial use.
9. Project Links
Related AI Model Articles

OpenMuse – CopilotKit Open-Source Personal AI Assistant
OpenMuse is an open-source personal AI assistant project developed by the CopilotKit team. Its core design philosophy is "giving an Agent a computer" — by combining a persistent browser, optional Linu...

In-Depth Review of Longcat-2.5-preview: Meituan's Next-Generation Multimodal Long-Range Agent Model
LongCat-2.5-preview is Meituan's latest next-generation large model. Building upon the 1.6T total parameters, approximately 48B active parameters, and native 1M token context of LongCat-2.0, it marks ...

Review of DeepSeek Harness Desktop: How the Official GUI Client Lowers the Bar for Agent Usage
DeepSeek Harness Desktop is the official graphical client launched by DeepSeek, designed to provide a visual interface for the originally command-line-based DeepSeek Harness framework. After users log...

Step Code – In-Depth Review of StepFun's Open-Source Terminal Programming Agent
Step Code is an open-source terminal programming agent launched by StepFun, licensed under the MIT License, which allows developers to complete the full workflow of code writing, debugging, execution,...
© All Rights Reserved. Some content on this site is partially generated by AI with human review.
