guizang-social-card-skill AI Skill

What it is

guizang-social-card-skill is an image-card Skill by Guizang (GitHub user op7418). The author says it's built for agent setups like Claude Code and Codex, and other local agents such as Cursor work too. Give it text, photos, video, screenshots, HTML or reference images, and it lays them out as Xiaohongshu (Rednote) cards, a two-image WeChat article cover set, or a Live Photo card.

Besides still cards, it can make Live Photo cards that move. It may fetch images, fonts and maps online and records where they came from in SOURCES.md; if you want extra images generated, it uses your agent's own model. It only makes images. It doesn't log in to any account or post. Guizang's dsh-guizang-design and guizang-material-illustration are also on the site; this entry covers the social cards only.

Who it's for:

People who make Xiaohongshu image posts and WeChat covers often and want them to look consistent.

Before you start:

Install with npx skills add, or git clone it into ~/.claude/skills/. You need Node.js, npm and Playwright. Live Photo cards also need ffmpeg, Python, uvx and makelive, and the Swift fallback only runs on macOS. It reads the text, images, video and locally installed fonts you give it; the author says local commercial fonts aren't uploaded. It uses curl to download images, isn't installed with curl piped to bash, and doesn't install hooks or background processes. Check the platform's rules before you post. AGPL-3.0 license; the LICENSE file in the repo is the reference. For white-label or other commercial deals, the repo has a separate document.

Based on the project's public documentation. Not tested by AI Inspiration World.