tua-orchestra

Coordinates parallel sub-agents by decomposing tasks into independent worker workstreams.

Updated Mar 7, 2026
One-click install
npx skills add https://github.com/shalevamin/Tau_agent --skill tua-orchestra-shalevamin
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: tua-orchestra
Source: https://github.com/shalevamin/Tau_agent/tree/main/tau-agent-main/skills/tua-orchestra
Command: npx skills add https://github.com/shalevamin/Tau_agent --skill tua-orchestra-shalevamin

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Complex tasks often exceed what a single linear agent pass can handle well, leading to shallow results or missed angles. This Skill splits broad jobs into focused workstreams executed by parallel worker agents, then merges their outputs centrally. ## Core Features & Use Cases - Parallel Worker Spawning: Uses sessions_spawn, sessions_send, and sessions_list to create, assign, and monitor multiple worker agents. - Task Decomposition Strategy: Breaks a job into independent workstreams such as UI, backend, tests, and research, each with a narrow objective and clear deliverable. - Routing and Synthesis: Routes workers to specialized skills like coding-agent when needed, while the conductor agent merges outputs and resolves conflicts. - Use Case: When asked to build a full-stack feature, spawn one worker for the UI, one for the backend, and one for tests, then synthesize their results into a coherent final deliverable. ## Quick Start Ask the agent to split this project into parallel workstreams and coordinate worker agents to complete each part, then merge the results.

Frequently Asked Questions about tua-orchestra

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run multiple AI agents in parallel on one task?

Decompose the task into independent workstreams, then spawn a worker per workstream using sessions_spawn and assign each a narrow objective via sessions_send. Poll progress with sessions_list and merge the outputs centrally.

What is the best way to split a coding project across sub-agents?

Assign one worker per concern, such as UI, backend or runtime, tests and verification, and research or docs. Give each worker minimal context, a clear deliverable, and the correct working directory.

When should I use subagents instead of spawning session workers?

Use the built-in subagents flow when a standard delegation pattern is enough. Spawn session workers when you need persistent, independently addressable agents that you can poll and re-task over time.

Why do parallel agents produce noisy or conflicting results?

Noise usually comes from overly broad prompts or duplicated large context blobs sent to every worker. Narrow each worker's objective, keep shared constraints identical, and re-run a worker with a tightened prompt if its first result is noisy.

Can a worker agent run Codex or Claude Code in a separate process?

Yes. When a worker should run Codex, Claude Code, Pi, or OpenCode as a separate process, activate the coding-agent skill for that worker while the conductor agent retains synthesis and quality control.