Harness Plugin
A team-architecture factory plugin for Claude Code that converts domain descriptions into coordinated agent teams using six pre-defined architectural patterns.
At a Glance
About Harness Plugin
Harness is a Claude Code plugin / skill that acts as a Team-Architecture Factory, transforming a plain-language domain description into a fully structured agent team with specialized agents and skills. It operates at the L3 Meta-Factory layer of the Claude Code ecosystem, generating agent definitions (.claude/agents/) and skill files (.claude/skills/) tailored to your project domain. Simply say "build a harness for this project" and Harness runs a six-phase workflow — from domain analysis through validation — to produce a ready-to-use multi-agent configuration. It supports a built-in evolution mechanism that feeds real-world deltas back into the factory for continuous improvement.
- Six Architecture Patterns — Choose from Pipeline, Fan-out/Fan-in, Expert Pool, Producer-Reviewer, Supervisor, and Hierarchical Delegation to match your task structure.
- Automatic Agent Definition Generation — Harness creates
.claude/agents/files for each specialist agent in your team, complete with roles and responsibilities. - Skill Generation with Progressive Disclosure — Auto-generates
.claude/skills/files that manage context efficiently, reducing token overhead across agent interactions. - Harness Evolution Mechanism — Use
/harness:evolveto capture deltas between the initial and shipped architecture, feeding improvements back into the factory for better next-generation drafts. - Two Execution Modes — Run in Agent Teams mode (default, using TeamCreate + SendMessage + TaskCreate) for multi-agent collaboration, or Subagents mode for one-off tasks.
- Validation & Testing — Built-in trigger verification, dry-run testing, and with-skill vs. without-skill comparison tests ensure generated harnesses work before deployment.
- Marketplace Installation — Install via
/plugin marketplace add revfactory/harnessand/plugin install harness@harness, or copy skills directly to~/.claude/skills/harness/. - Multilingual Trigger Support — Activate Harness in English, Korean (하네스 구성해줘), or Japanese (ハーネスを構成して).
- Broad Use Case Coverage — Includes reference examples for deep research, full-stack development, code review, data pipelines, marketing campaigns, and more.
Community Discussions
Be the first to start a conversation about Harness Plugin
Share your experience with Harness Plugin, ask questions, or help others learn from your insights.
Pricing
Open Source
Fully free and open-source under the Apache License 2.0. Free to use, modify, and distribute.
- Six team architecture patterns
- Automatic agent and skill generation
- Harness Evolution Mechanism
- Agent Teams and Subagents execution modes
- Marketplace and direct installation
Capabilities
Key Features
- Six team architecture patterns: Pipeline, Fan-out/Fan-in, Expert Pool, Producer-Reviewer, Supervisor, Hierarchical Delegation
- Automatic agent definition generation (.claude/agents/)
- Skill generation with Progressive Disclosure for context management
- Harness Evolution Mechanism via /harness:evolve
- Agent Teams and Subagents execution modes
- Trigger verification, dry-run testing, and comparison tests
- Marketplace and direct installation support
- Multilingual trigger support (English, Korean, Japanese)
- Six-phase workflow: Domain Analysis → Team Design → Agent Definition → Skill Generation → Integration → Validation
