Multi-Agent System Orchestrator

What it does

Enables the design and coordination of complex multi-agent systems where a central Orchestrator manages the lifecycle of specialized worker agents. It focuses on "knowledge harvesting," allowing the system to extract learnings from completed tasks and store them in a vector database for future agents to use via RAG.

When to use it

Use this skill when building autonomous agent projects (like the PSI Engine) that require:

• Spawning and monitoring multiple agent processes via PTY.
• Implementing a learning loop where task outputs are distilled into patterns.
• Integrating ChromaDB for long-term agent memory and context injection.
• Coordinating communication and state sharing between different AI agents.

What's included

• Instructions: Detailed Python implementations for spawning agents, non-blocking output monitoring, and knowledge harvesting logic.
• Architecture: Visual and logic flows for Orchestrator-Worker-Knowledge Base interactions.
• Troubleshooting: A decision tree and common issues table for PTY buffer management and embedding tuning.

Compatible agents

Suitable for high-level orchestrator agents capable of executing Python and managing subprocesses, typically within a Codex or Claude-like environment.