harness

Orchestrates agent teams and skills to automate reproducible workflows via CLAUDE.md pointers.

Updated Apr 21, 2026

Search Tags:#automation #skills #agents #orchestration #governance #harness #team-design

Ready to run

npx skills add https://github.com/hobbong21/nuxia2 --skill harness-hobbong21

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: harness
Source: https://github.com/hobbong21/nuxia2/tree/main/.claude/skills/harness
Command: npx skills add https://github.com/hobbong21/nuxia2 --skill harness-hobbong21

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Harness orchestrates the creation and evolution of agent teams and their skills to build reproducible automation harnesses.

Core Features & Use Cases

Defines domain-specific agent teams and the skills they will use to automate complex workflows.
Provides templates and governance for Phase-based orchestration, drift checks, and maintenance.
Enables long-lived collaboration by persisting agent/skill definitions and CLAUDE pointers.

Quick Start

Create your first agent under .claude/agents and its initial skill under .claude/skills, then register a CLAUDE.md harness pointer and run Phase 0 to begin.

Frequently Asked Questions about harness

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

FAQPage Schema

How do I orchestrate agent teams for scalable automation workflows?▼

To orchestrate agent teams for scalable automation, you define domain-specific agents under .claude/agents and their skills under .claude/skills, then register a CLAUDE.md harness pointer to manage reproducible execution and governance workflows.

What is an automation harness and when do I need one for agent orchestration?▼

An automation harness is a structured framework that orchestrates the creation and evolution of agent teams and their skills. You need one when automating complex, long-lived workflows that require reproducibility, drift checks, and ongoing governance.

How do I set up agent teams and skills using a .claude directory structure?▼

To set up agent teams, create agent definitions in the .claude/agents directory and initial skills in the .claude/skills directory. Register a CLAUDE.md harness pointer to link them, then run Phase 0 to begin orchestration.

Does agent orchestration require a specific repository context to function?▼

Yes, agent orchestration requires a repository context containing .claude/agents and .claude/skills definitions. A CLAUDE.md pointer is also required, while optional references can be added to support scalable collaboration.

How do I maintain governance and prevent drift in automated agent teams?▼

You maintain governance and prevent drift in automated agent teams by applying the provided templates for Phase-based orchestration, running drift checks, and executing ongoing maintenance workflows defined within the harness.

Can I use this orchestration approach for long-lived collaboration across multiple agents?▼

Yes, this approach enables long-lived collaboration by persisting agent and skill definitions alongside CLAUDE.md pointers, ensuring that automated workflows remain reproducible and governed over time.