AI Agent Writes and Publishes to WeChat Official Account: A Full Walkthrough with Zhiyu and baoyu-skills

This walkthrough shows that AI Agents are moving beyond chat into real production workflows — writing, image generation, and CMS publishing — all orchestrated from a single prompt. For Western developers, it signals that open-source skill ecosystems like baoyu-skills are becoming platform-agnostic building blocks, lowering the barrier to building autonomous content pipelines.

The open-source AI Agent Zhiyu (zhitalk.chat), combined with the baoyu-skills writing toolkit, now handles the entire content pipeline for WeChat Official Accounts: writing, image generation, and draft publishing. A developer demonstrates the full flow by installing Zhiyu globally via npm, configuring it with a Kimi model, and adding baoyu-skills as a skill set.

The test starts simple: feeding a local text file and image, then asking the agent to publish. The agent loads the baoyu-post-to-wechat skill, reads the files, and runs a Bun script to push a draft to WeChat. Next, image generation is added via Alibaba Cloud's DASHSCOPE API — the agent prompts the user for confirmation, then calls baoyu-image-gen to produce a cover image.

In the final end-to-end test, the developer asks for a 2000-word article on "Which industries will AI Agents change?" with 2-3 inline images and a cover. The agent writes the article, generates images, and publishes. The first attempt misses inline images — but a follow-up command "check again" triggers the agent to fix the issue and re-publish successfully. The entire workflow runs on standard, reusable skills that work across multiple AI Agent platforms.