What is stick-breaking modality selection?

It is a mathematical method using Poisson-Dirichlet weights to determine which input stream—such as audio, video, or data—should receive the system's primary attention at any given time.

Why does the skill output observations as s-expressions?

S-expressions provide a structured, symbolic format that is ideal for categorical processing, allowing other tools to easily parse and act upon coaching events.

How does the 'The Queen' voice persona function?

The persona uses a high-quality British English voice (Serena Premium) to provide persistent, authoritative coaching guidance throughout a LiveKit session.

What is the purpose of the ε-machine in this skill?

The ε-machine acts as a dynamic sufficiency gate, ensuring the AI does not attempt a task until it has verified that all necessary domain-specific skills are currently loaded.

LiveKit Omni-modal Coach

Name: LiveKit Omni-modal Coach
Author: plurigrid

byplurigrid

•

Data Science & ML

Facilitates real-time, multi-modal continuous coaching and guidance through LiveKit using dynamic modality selection and symbolic expression processing.

The LiveKit Omni-modal Coach skill integrates Claude with real-time WebRTC infrastructure to provide a sophisticated multi-modal coaching environment. It features continuous listening and guidance via 'The Queen' voice persona, employing stick-breaking modality selection (Poisson-Dirichlet weights) to intelligently prioritize audio, video, and data streams. By utilizing an ε-machine for dynamic sufficiency gating, the skill ensures that actions are only performed when prerequisite skills are verified. All observations are formatted as s-expressions, enabling seamless integration with categorical processing workflows and complex AI orchestration systems.

Key Features

01Symbolic s-expression output for structured categorical processing

022 GitHub stars

03Continuous multi-modal coaching across audio, video, and data tracks

04Dynamic sufficiency gating via ε-machine skill verification

05GF(3) role conservation for stable multi-agent coordination

06Stick-breaking modality selection for prioritized attention management

Use Cases

01Real-time pair programming and technical guidance sessions

02Building self-verifying AI agents with dynamic skill-loading requirements

03Automated multi-modal monitoring and documentation of live coding

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add plurigrid/asi livekit-omnimodal

For use in Claude.ai and ChatGPT

Download Skill