Viral Content Pre-Check (Paid Version)

The Problem It Solves

Video content creators face three major pain points before publishing: compliance risks, uncertain audience reactions, and difficulty predicting viral potential. Manual reviews are inefficient, often missing sensitive terms or overlooking advertising law violations. The real response of target audiences remains unknown in advance, and whether content will go viral relies heavily on experience. These uncertainties can lead to content being taken down for violations, poor engagement, or legal issues.

How the Skill Works

Core Capabilities

The skill offers three independent modules that can be used individually or combined:
- Compliance Detection (the compliance module) matches sensitive words and advertising law provisions using an authoritative database, with LLM-powered rewriting suggestions for second-level accuracy.
- Persona Feedback & User Comments (the persona module) simulates comments and reactions from different user groups (e.g., age, gender, occupation, consumption tier) via LLM, providing auxiliary references in seconds.
- Viral Probability Prediction (the burst module) scores content using LLM to forecast its viral potential, serving as a quick assessment tool.

Key Workflow

The skill uses an asynchronous unified task interface to ensure efficiency without interrupting sessions:
1. Task Creation: Users submit the text to check, specify the publishing platform (e.g., video channel or WeChat), and select modules. The skill immediately returns a task_id via the POST /tasks endpoint.
2. Progress Polling: Multiple short polls query the task status via GET /tasks/{id}, avoiding synchronous blocking. Each poll updates the user on progress until completion or failure.
3. Report Generation: Upon success, Markdown content from each module's results is extracted and merged into a complete report, including compliance conclusions, simulated user feedback, and viral scores. The report is saved as a file.

Applicability and Considerations

• Text-Only Analysis: The skill only analyzes text content (e.g., voiceover scripts, titles, subtitle bars) and does not process video visuals or audio.
• Limitations of Compliance Detection: The database, while updatable, may lag behind platform rules or legal changes; manual review is recommended for critical publications.
• Nature of Simulated References: Persona feedback and viral predictions are based on LLM simulations, intended as auxiliary references that do not guarantee accuracy and should not be the sole basis for publishing decisions.
• Cost and Execution: Task execution consumes points (estimated around 38 points, actual deduction determined server-side), implemented via asynchronous polling. User confirmation is required before proceeding; points are refunded on failure with retry options.