Viral Script Generator for WeChat Channels

The Problem: Crafting Scripts That Actually Go Viral

In the WeChat Channels (视频号) ecosystem, the first 3 seconds of an oral video, the script structure, and the conversational tone directly determine watch-through rates and sharing potential. Many content creators struggle to conceive a script from scratch that aligns with the platform's style while being both engaging and conversion-oriented. Common pitfalls include hitting creative blocks or producing scripts full of advertisement-like language and loose logic that fail to gain algorithmic traction.

How It Works: A Systematic Workflow from Input to Deliverable

This skill (@org-28ib33ph/lingyi-wx-viral-script-generator) systematizes and standardizes the script creation process. Crucially, it does not embed any LLM itself. It relies entirely on the model currently configured in your Agent to generate the creative core, while the skill's scripts handle structural validation and final rendering to ensure robust output.

The workflow is a strict six-step pipeline:

1. Information Collection with Smart Fallbacks: It requests five key fields from the user: creation_purpose (e.g., product seeding), industry, target_audience, topic, and product (optional). If information is missing, it prompts the user once for all missing items. If still unresolved, it performs a fallback generation based on an internal "viral potential checklist," ensuring the script follows a proven logic structure.
2. Intelligent Script Type Selection: Based on the collected info, it automatically selects the most suitable type from six categories like pain_point, scene, or drama, and defines corresponding campaign_types (e.g., short video seeding, brand exposure).
3. Schema-Driven Structured Generation: This is the core creative step. It strictly adheres to a predefined JSON output-schema, populating all fields including title, duration_sec, structure.hook (opening hook), structure.body.segments (body paragraphs), and structure.cta (call-to-action, index fixed at 99). Each segment requires separate visual (on-screen description) and dialogue (spoken lines) fields.
4. Rigorous Local Compliance Check: Immediately after generation, the script undergoes multiple checks: scanning and automatically replacing extreme terms (e.g., "best", "first"), verifying medical/efficacy claim redlines, ensuring price mentions include specifications, and optimizing conversational naturalness (e.g., sentence length ≤15 characters, using "..." for pauses, avoiding书面 connectors). It will rewrite if non-compliant (up to 3 rounds).
5. Validation, Rendering, and Delivery: The JSON data is passed to a dedicated Python script, render_script.py. This script performs a final schema compliance check. Upon passing, it renders everything into a single, cleanly formatted Markdown document with a summary table and a full-text appendix. This Markdown file is the sole deliverable; temporary JSON files are cleaned up.
6. Async Server Submission (Non-blocking): Concurrently with delivery, the skill asynchronously submits the result to a server for logging and analysis. This is a non-blocking operation; any network failure does not prevent you from receiving the final script document.

Key Considerations and Boundaries

• Model Dependency: The creativity and textual quality of the script depend entirely on the LLM configured in your Agent. The skill provides the structure, workflow, and quality assurance framework. Ensure your model is capable of understanding and filling all fields.
• Garbage In, Garbage Out: The more specific and insightful the provided creation_purpose, industry, audience, and topic, the more targeted and potentially viral the generated script will be.
• Local & Model-Agnostic: The entire process runs in your local environment. No additional API keys are required for this specific skill.
• Version and Scope: This is an experience version (v0.3.0), focused solely on generating WeChat Channels oral scripts. It is not a video editing or publishing tool. It generates a structured, ready-to-use copy blueprint.