ZeroOne Tech AI-Powered Article Generation for Xiaolvshu & WeChat Official Accounts

The Problem It Solves

Creating "viral" graphic posts for platforms like Xiaolvshu (WeChat Moments) or Official Accounts is a multi-step process that demands both copywriting and aesthetic skills. It involves ideation, designing cover and inset images, writing compelling copy, and crafting multiple title candidates. This manual workflow is time-consuming and makes consistency difficult. The Article Generation skill addresses this "cold start" challenge by transforming a set of core inputs into a complete, editable Markdown draft with AI-generated visuals, streamlining the path from concept to first draft.

How It Works: Asynchronous Generation & Delivery

The skill's core is an asynchronous task interface, which operates differently from a typical synchronous API call.

1. Information Collection & Task Creation: After the user provides essential inputs (e.g., topic --text, product name --product-name, selling points, audience description), the skill calls the POST /api/v1/content/article-generation endpoint. This immediately creates an async task and returns a task_id and initial status, not the final result. Generating the full report with multiple images requires server-side processing time (typically 1-2 minutes).
2. Polling for Progress & Results: With the task_id, the skill must continuously poll the GET /api/v1/content/article-generation/{id} endpoint to check status. This design prevents timeouts on long-running operations. Each poll returns the current status (e.g., processing) and a progress percentage. Once the status is completed, the final report is fetched from the data.markdown field in the response.
3. Delivery & Presentation: The delivered Markdown report is pre-rendered by the backend, containing the H1 title, image embeds (![alt](url)), body text, and candidate titles. The skill renders this Markdown content directly for the user to see, states the points consumed (data.total_points), and saves the complete .md file locally.

Key Features & Considerations

• Content Formats: It supports two primary modes: Xiaolvshu (compact, default 3:4 images) and Official Accounts (fixed 2.35:1 cover + separate inset images). The structure of the output Markdown is fixed and should be used as-is; reorganizing it locally is discouraged.
• Billing & Retry Mechanism: Each successful task incurs a point cost (starting around 60 points). This skill has no /retry endpoint. If a task fails, you must recreate a completely new task with all original parameters, which will be billed independently.
• Dynamic Tone Configuration: The "content tone" (e.g., "knowledge-based") is optional. Its available values are fetched dynamically via the GET /api/v1/content/config endpoint. Hardcoding example values in the script is strictly forbidden to avoid breakage if the backend enum updates.
• Output as a Draft: The cover images, inset images, and copy are AI-generated and may contain factual inaccuracies or brand inconsistencies. All content must undergo manual review before publication to ensure accuracy and compliance.