convez.ai
SkillDevOps & InfraestruturaAvançado
theopengroup

Rune Harness

theopengroup/rune-harness

Harness de IA para desenvolvimento sistemático com multi-agent collaboration

Ver no GitHub
InicialPluginScore · de 0 a 100

Plugin Claude Code que guia equipes de desenvolvimento na transformação sistemática do uso de IA, construindo CLAUDE.md, bibliotecas de Skills e workflows de agentes. Funciona através de análise do codebase, propostas baseadas em evidências e confirmação deliberada de cada decisão, evitando geração cega de arquivos.

Instalação

$npx convez-ai theopengroup/rune-harness

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
1
Forks
0
Acompanhando
0
Saúde
28%
Último commit
há 7 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Estabelecer convenções e padrões de IA em projetos Go/Java/Python
  • Construir e manter CLAUDE.md e bibliotecas de Skills organizadas
  • Implementar workflows de agentes para CI/CD e qualidade de código

Veredito

Ferramenta sólida para equipes que buscam profissionalizar o uso de IA em desenvolvimento, estruturando harnesses completos com governance clara. Requer investimento inicial significativo mas compensa em projetos grandes com múltiplos agentes.

README

A tradução para o português deste README ainda está na fila.

rune-harness

Forge your AI development harness, one rune at a time.

A Claude Code plugin that guides development teams through AI-assisted development transformation via multi-agent collaboration — systematically building CLAUDE.md, Skills libraries, and agent workflows to establish a complete AI development harness for your project.

中文 | English


English

What is rune-harness?

rune-harness is a Claude Code plugin that helps software teams move from ad-hoc AI usage to a systematic, compounding AI development practice. Instead of generating files blindly, it interviews your codebase and your team, then proposes each convention one at a time for your confirmation.

Core principles:

  • Harness First — Build constraints + tools + feedback loops, not just documentation
  • Confirm, Don't Generate — AI analyzes evidence → proposes → developer confirms. Every output is a deliberate decision
  • Compounding Improvement — Each improvement prevents a class of future failures

Installation

# Clone to Claude Code plugins directory
git clone https://github.com/theopengroup/rune-harness.git ~/.claude/plugins/rune-harness

# Or install via marketplace (after configuration)
# /plugin install rune@rune-harness-marketplace

Commands

CommandDescription
/runeLaunch the Rune transformation wizard (first-time setup)
/rune-statusView current maturity level and coverage dashboard
/rune-updateContinuously improve your harness based on feedback

Language Support

LanguageStatus
Go✅ Full support
Java / Spring Boot🔜 Planned
Python🔜 Planned

中文

rune-harness 是什么?

rune-harness 是一个 Claude Code 插件,通过多 Agent 协作,引导开发团队系统化完成 AI 辅助开发转型——从 CLAUDE.md 编写、Skills 库建设到子 Agent 工作流设计,逐步建立属于自己项目的 AI 开发 Harness。

安装

git clone https://github.com/theopengroup/rune-harness.git ~/.claude/plugins/rune-harness

使用命令

命令说明
/rune启动 Rune 转型向导(首次使用)
/rune-status查看当前成熟度与覆盖度看板
/rune-update基于近期使用反馈持续优化 Harness

Plugin Architecture

rune-harness/
├── skills/            # User-facing skill entry points
├── agents/            # Multi-agent role definitions
├── skills-library/    # Installable skills catalog
│   ├── universal/     # Language-agnostic skills
│   └── golang/        # Go-specific skills
├── hooks/             # Lifecycle hooks
└── templates/         # CLAUDE.md generation templates

Multi-Agent Design

AgentRole
transformation-plannerScans project, builds roadmap, coordinates specialists
dev-transformerCLAUDE.md + Skills + development workflows
test-transformerTest strategy + TDD workflow + test harness
ops-transformerCI/CD integration + deployment workflows
go-scannerDeep analysis of Go project structure and conventions
spec-reviewerSpec compliance review after each implementation
code-quality-reviewerCode quality review with >80% confidence threshold

Rune Maturity Levels

LevelNameAI Code Contribution
L1Aware< 10%
L2Tooled10–25%
L3Systematic25–40%
L4Workflow-driven40–60%
L5Intelligent60%+

License

MIT

Ver o original no GitHub

Bom para

Relacionados