Conversational WeChat Writing Assistant

Problem

Writing a WeChat article often gets stuck on two issues: the opening lacks a hook, and the title or subtitles read like a table of contents rather than strong claims. LLM drafts can also sound formulaic, like documentation. This skill breaks writing into confirmable steps, reducing off-topic and vague output from one-shot long-form generation.

How It Works

It defaults to WeChat, with Xiaohongshu as an optional target, and reads the matching references style file before running. On first use, it initializes ip-info.md by asking for the author name, a one-line identity, and closing bio. The workflow follows topic confirmation -> main title -> subtitle framework -> hook sentence -> drafting -> quality self-check -> AI-tone polishing, with user confirmation at each step. Built-in polishing removes or replaces high-frequency AI phrases such as “in addition,” “it is worth noting,” and “deeply explore,” controls em dashes, ellipses, and parallel patterns, then outputs content-creation/--article.md with an English image prompt.

Boundaries

It does not generate images automatically and does not replace deep research. If the topic lacks source material, provide conversation logs, keywords, or a clear thesis first. It is best suited for turning existing ideas into a more personal WeChat draft, not for content requiring strict citation, multi-pass fact-checking, or brand-compliance review.