aiworkskills/wechat-article-skills

What it solves

This project automates the repetitive labor involved in daily WeChat Official Account operations. It handles the entire pipeline from topic selection and writing to review, formatting, image generation, and publishing to the draft box, reducing the need for design or coding skills.

How it works

It is a set of AI Agent Skills (based on the OpenClaw standard) that can be integrated into AI tools like Claude Code, Cursor, or the Claw series. The system uses a modular approach with 9 specialized skills:

• Main Process: Orchestrates the entire workflow.
• Topic Selection: Researches hot topics and recommends options.
• Writing: Generates content based on specific style guides.
• Review: Checks for sensitive words, typos, and "AI-sounding" prose.
• Formatting: Converts Markdown to WeChat HTML using various templates and layout components.
• Images: Generates covers and in-text images using a method-driven approach (e.g., avoiding AI clichés).
• Publishing: Uses the WeChat API to send content directly to the draft box.
• Stickers: Handles multi-image push processes.
• Assets: Manages a business knowledge base and configuration presets via .aws files.

Users can configure their account positioning, writing style, and visual preferences through a visual configuration platform (aiworkskills.cn) or via YAML files.

Who it’s for

Individuals and teams operating WeChat Official Accounts who want to automate repetitive tasks while maintaining human-in-the-loop control, as the agent pauses for confirmation at every step.

Highlights

• Human-in-the-loop: Designed to stop for user confirmation at every stage, allowing for interruptions and modifications.
• Method-driven imagery: Uses a seven-step derivation process for covers and specific criteria for in-text images to ensure high-quality, non-generic visuals.
• Visual Layout Components: Includes 9 specialized layout components (e.g., gold-sentence cards, data cards) to handle structural elements that standard HTML cannot.
• Broad Ecosystem Support: Compatible with 13+ Claw-series tools and mainstream agents like Cursor and Claude Code.
• Local Privacy: API keys and WeChat credentials are stored locally in aws.env and are not uploaded to any server.