Ark WeChat MP Operation AI Assistant

Problem

WeChat Official Account writing often has conflicting constraints: data-heavy sections can read like walls of text, images must not be forced at fixed intervals, abstracts have strict length limits, and sensitive or AI-flavored wording needs compliance cleanup. This skill turns a topic into a publishable draft pipeline covering writing, image planning, summarization, compliance replacement, and draft-box output.

How It Works

• Writing and image plan: Given a topic, it generates the article and emits illust_plan.txt. A script locates image slots by marker text; if markers are missing, it falls back to even distribution.
• Content-driven illustration: Images are decided by scene rather than a fixed cadence. Narrative sections may receive more images, while data-heavy sections should stay table-first. The materials mention parallel image generation via mxai.
• Data tables: Factual metrics such as customer value, revenue, and share can be rendered as Markdown tables for faster scanning.
• Abstract and compliance: The abstract is extracted from the first 120 characters, preferring breaks at ., !, or ?, and capped at 54 characters. Compliance replacement is context-aware, e.g. ChatGPT/GPT becomes “smart assistant,” and AI-generated becomes “generated,” with a 50% probability of removing some AI-flavored terms.
• Draft output: The final result goes to the Official Account draft box for human review before publishing.

Boundaries

It is useful for turning a clear topic into a reviewable draft. It is not a fact-checker or final legal-compliance decision. Table quality, abstract quality, and compliance replacements all depend on the source material and still need editorial review, tone checking, and confirmation before publishing.