XHS Visual Director

XHS Visual Director is an Agent Skill by GitHub user ziguishian for planning and generating Xiaohongshu image posts. You answer its ten questions to get your topic straight. It then lays out a 6 to 8 page structure in one consistent style and makes a single visual confirmation image for you to check. Only after you approve does your AI image tool produce the full set, along with the post copy. Tools that go straight to images skip this step; here the style is settled on one test image before the full set is made. The repo also includes docs, templates and examples.

Who it's for:Creators and social media managers who make Xiaohongshu image posts and want every set to look consistent.

Before you start: The README only covers Codex: put the repo's skill folder into your Codex Skills directory, with skill/SKILL.md and agents/openai.yaml as the entry points. It doesn't say how to install it in Claude Code or OpenClaw.

It doesn't sign in to Xiaohongshu, store account credentials, or post or interact for you, and it runs nothing in the background. The topics, drafts, product photos and reference images you give it go to the AI model you use, which also makes the images. Costs and data retention depend on that service; the author doesn't say.

The demo screenshots in the repo show third-party logos and comment avatars, and there's no separate license note for them. MIT license.

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