youtube_moderation_prototype

This moderation prototype is designed to maintain a healthy and safe YouTube community environment by providing a structured, rule-based approach to chat filtering. It acts as a testing scaffold, allowing developers to validate pattern fidelity and classification accuracy before porting the logic to native Qwen or Gemma environments. By implementing rigorous validation steps, the skill ensures high precision in identifying disruptive behaviors while maintaining low false-positive rates for legitimate user interactions.

• CAPS SPAM CHECK: Identifies and blocks messages exceeding length thresholds that contain a high ratio of uppercase characters.

• REPETITION CHECK: Monitors the chat history to detect and mitigate spamming behaviors through repeated, exact-match phrases.

• RATE LIMIT CHECK: Enforces flow control by warning or blocking users who exceed defined message frequency limits within a 30-second window.

• TOXIC CONTENT CHECK: Scans incoming messages against a configurable list of keywords to identify and block harmful, toxic, or offensive language with high confidence.

• LEGITIMATE MESSAGE ROUTING: Automatically identifies safe content and routes it to banter or engagement services, ensuring seamless community interaction.

• The skill requires a high-fidelity evaluation (>= 90%) against a pre-defined benchmark set of 110 test cases including spam, toxicity, and legitimate dialogue.

• It is architected for integration with the 0102 agent swarm framework, enabling autonomous moderation across multiple platforms.

• Inputs include live chat streams; outputs consist of structured JSON logs identifying the decision (block/allow/warn), the reason, and a confidence score for auditing.

• This module is strictly a prototype meant for testing and validation; it should be integrated into broader WSP (Windsurf Recursive Engine) workflows for production-grade deployment.

• Users should ensure toxic_patterns.json is correctly loaded to manage fuzzy and exact keyword matching for evolving moderation requirements.