orchestration

Coordinate isolated sub-agents across linear, parallel, DAG, and iterative loop topologies.

18|3|Updated Jan 12, 2026
One-click install
npx skills add https://github.com/buYoung/skills --skill orchestration-buyoung
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: orchestration
Source: https://github.com/buYoung/skills/tree/main/skills/orchestration
Command: npx skills add https://github.com/buYoung/skills --skill orchestration-buyoung

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Coordinates long-running, multi-phase tasks by tying together a manager orchestrator with isolated sub-agents that execute distinct phases without leaking context. It provides run-id based resumability, strict per-phase isolation, and a framework for combining heterogeneous shapes such as linear pipelines, parallel fan-outs, dependency DAGs, iterative loops, and recursive decomposition.

Core Features & Use Cases

  • Manager orchestrator decomposes tasks into heterogeneous phases (investigate, produce, run, evaluate, analyze, report) and routes work to phase sub-agents.
  • Per-phase isolation using per-run directories and file-path communication, enabling resumability and clean lifecycle management.
  • Topologies: linear pipeline, parallel fan-out, dependency DAG, iterative loop, and recursive nesting to fit real dependency structures.
  • Explicit halts: go/no-go before expensive steps and review accept/reject for user judgment to preserve long-autonomous runs.
  • Run-id based versioning, change_log, and resume semantics to safely adapt to changing conditions.
  • Reference templates and governance guidance in references/ for phase prompts, tool-permission defaults, and platform notes.

Quick Start

Initiate a run that decomposes a complex task into phased sub-agents and runs each phase in isolation.

Frequently Asked Questions about orchestration

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

FAQPage Schema
How do I coordinate long-running multi-phase tasks without leaking context between isolated sub-agents?

You coordinate long-running multi-phase tasks using a manager orchestrator that decomposes work into isolated sub-agents. Each phase executes within per-run directories, communicating strictly through file paths to prevent context leakage and ensure clean lifecycle management.

Can I resume a halted pipeline run if conditions change midway through task execution?

Yes, you can resume a halted pipeline run because the orchestration framework enforces run-id based versioning and maintains a change_log. This resumability semantics allows the system to safely adapt to changing conditions after a controlled halt.

What topology should I use to structure complex sub-agent workflows with dependency DAGs and parallel fan-outs?

You structure complex sub-agent workflows using supported topologies including linear pipelines, parallel fan-outs, dependency DAGs, iterative loops, and recursive nesting. These heterogeneous shapes fit real dependency structures to match your specific workflow requirements.

How do I implement go/no-go halts before expensive steps in an autonomous agent pipeline?

You implement go/no-go halts by configuring explicit halts before expensive steps within the manager orchestrator. The framework also supports review accept/reject checkpoints for user judgment, preserving long-autonomous runs by preventing unchecked costly operations.

Do I need reference templates to set up phase prompts and tool permissions for sub-agents?

You need reference templates to quickly set up sub-agents, as the framework provides governance guidance in references for phase prompts, tool-permission defaults, and platform notes. These templates accelerate adoption and ensure correct isolation configurations.