Aria — AI-DDD para Desenvolvimento Colaborativo
10CG/Aria
Metodologia AI-DDD para IA como colaboradora genuína em software
Ver no GitHubPlugin que implementa a metodologia AI-DDD (AI-Assisted Domain-Driven Design) para transformar Claude em um colaborador genuíno em projetos de software. Estrutura o desenvolvimento através de um ciclo de 10 passos, mantendo estado do projeto, sincronizando documentação e especificações, e garantindo que IA compreenda o contexto e as convenções do seu projeto.
Instalação
npx convez-ai 10CG/Aria1 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
Casos de uso
- Estruturar fluxo de trabalho entre desenvolvedores e Claude com contexto persistente
- Especificar requisitos com OpenSpec e rastrear mudanças com auditoria
- Desenvolver com TDD garantindo testes primeiro e documentação sincronizada
- Brainstorming colaborativo onde IA participa de decisões arquiteturais
Veredito

Aria é uma abordagem séria e estruturada para elevar IA de 'ferramenta que gera código' para 'colaboradora que entende projeto'. Ideal para equipes que buscam reprodutibilidade, rastreabilidade e contexto persistente—exige disciplina metodológica mas entrega ganhos reais em consistência e eficiência.
README
A tradução para o português deste README ainda está na fila.
Aria
Make AI a genuine collaborator in your software projects
What is Aria?
Aria is an AI-DDD (AI-Assisted Domain-Driven Design) methodology that enables AI assistants like Claude Code to deeply participate in the entire software development lifecycle through structured workflows.
Unlike traditional "AI writes code" tools, Aria focuses on: how to make AI understand project intent and become a valuable collaborator.
| Traditional Mode | Aria Mode |
|---|---|
| AI is a tool — you ask, AI answers | AI is a collaborator — AI understands, you confirm, you deliver together |
Aria 2.0 (v2.0.0, in progress) extends the methodology to autonomous execution. See docs/architecture/system-architecture.md for the two-layer architecture.
Why Aria?
The Problem
- AI suggestions don't follow your project conventions
- You re-explain project context every session
- Code and documentation drift apart
- Requirement changes have no audit trail
The Solution
| Feature | Description |
|---|---|
| State Awareness | AI automatically scans your project and understands current progress |
| Spec First | OpenSpec standardizes requirement descriptions |
| Ten-Step Cycle | Structured AI collaboration workflow |
| Docs in Sync | Architecture docs evolve with your code |
| TDD Driven | Test-first development with enforcement |
| Collaborative Thinking | Structured brainstorming with AI participation |
See the Aria 2.0 PRD for the roadmap and autonomous execution vision.
Quick Start
Prerequisites
- Claude Code installed and authenticated
- Git 2.x+ (if using the standards submodule)
Install the Aria Plugin
# Add marketplace
/plugin marketplace add 10CG/aria-plugin
# Install (Skills + Agents included)
/plugin install aria@10CG-aria-plugin
Install Standards (Optional)
The standards submodule provides OpenSpec requirement specifications. Skip this if you don't need spec-driven workflows.
# HTTPS
git submodule add https://github.com/10CG/aria-standards.git standards
# Or SSH
git submodule add git@github.com:10CG/aria-standards.git standards
Configure Your Project
Create .aria/config.json from the template, or simply start using Aria:
# Scan project status
/aria:state-scanner
# Create a requirement spec
/aria:spec-drafter
# Structured brainstorming
/aria:brainstorm
# Call a specialized agent
/aria:tech-lead Please plan the architecture for this feature
How It Works: The Ten-Step Cycle
flowchart LR
subgraph A["A. Planning"]
A0[A.0 Scan] --> A1[A.1 Spec] --> A2[A.2 Plan] --> A3[A.3 Assign]
end
subgraph B["B. Development"]
B1[B.1 Branch] --> B2[B.2 Develop + Review]
end
subgraph C["C. Integration"]
C1[C.1 Commit] --> C2[C.2 Merge]
end
subgraph D["D. Closure"]
D1[D.1 Update] --> D2[D.2 Archive]
end
A3 --> B1
B2 --> C1
C2 --> D1
Each phase has a dedicated Skill that ensures consistent, repeatable workflows:
| Phase | What Happens |
|---|---|
| A. Planning | Scan project state → Create spec → Break down tasks → Assign agents |
| B. Development | Create branch → Develop with TDD + code review |
| C. Integration | Generate commit message → Merge to main |
| D. Closure | Update progress → Archive spec |
What You Get
Skills (34 user-facing + 7 internal = 41 total)
| Category | Skills | Purpose |
|---|---|---|
| Cycle Core | state-scanner, workflow-runner, phase-a-planner, phase-b-developer, phase-c-integrator, phase-d-closer, spec-drafter, task-planner, progress-updater | Structured ten-step workflow |
| Collaborative Thinking | brainstorm | Structured brainstorming sessions |
| Git Workflow | commit-msg-generator, strategic-commit-orchestrator, branch-manager, branch-finisher | Commit and branch management |
| Dev Tools | subagent-driver, tdd-enforcer, requesting-code-review | TDD enforcement, code review |
| Architecture Docs | arch-search, arch-update, arch-scaffolder, api-doc-generator | Keep docs in sync with code |
| Requirements & Issues | requirements-validator, requirements-sync, forgejo-sync, openspec-archive, issue-triage | Requirement tracking and issue triage |
| Project Adaptation | project-analyzer, agent-gap-analyzer, agent-creator | Analyze projects, identify Agent gaps, generate configs |
| Observability & Estimation | aria-context-monitor, ai-native-estimator, aria-dashboard | Context/token telemetry, workload estimation, progress visualization |
| Feedback & Diagnostics | aria-report, aria-doctor | Bug reports and environment health checks |
| Infrastructure (7 internal, non-user-invocable) | config-loader, audit-engine, agent-team-audit, agent-router, arch-common, git-remote-helper, aria-token-telemetry | Configuration, audit orchestration, task routing, shared infra |
Agents (11)
| Agent | Role |
|---|---|
| tech-lead | Technical decisions and architecture planning |
| context-manager | Cross-agent context management |
| knowledge-manager | Knowledge base management |
| code-reviewer | Code review |
| backend-architect | Backend architecture design |
| mobile-developer | Mobile development |
| qa-engineer | Quality assurance |
| ai-engineer | AI/LLM application development |
| api-documenter | API documentation |
| ui-ux-designer | Interface design |
| legal-advisor | Legal and compliance documentation |
Use Cases
| Scenario | How Aria Helps |
|---|---|
| New Feature | End-to-end flow from requirements to code |
| Bug Fix | TDD-driven fix workflow |
| Refactoring | Code evolution with architecture docs in sync |
| Code Review | Automated convention compliance checks |
| Knowledge Transfer | Help newcomers understand a project quickly |
| Technical Decisions | Structured brainstorming and solution design |
OpenSpec: Requirement Specifications
A standardized format for describing requirements so AI and humans agree on "what to build":
| Level | When to Use | Output |
|---|---|---|
| 1 (Skip) | Simple fixes | No spec needed |
| 2 (Minimal) | Medium features | proposal.md |
| 3 (Full) | Architecture changes | proposal.md + tasks.md |
Aria plugin reads specs from openspec/changes/ in your project root directory (not inside standards/). The standards submodule provides the methodology definitions that the plugin references.
Project Structure
Your project (after adopting Aria):
your-project/
├── .aria/
│ └── config.json # Project configuration
├── openspec/
│ └── changes/ # Your requirement specs go here
├── standards/ # (Optional) Methodology specs submodule
├── docs/ # (Recommended) Architecture docs
│ └── architecture/ # Kept in sync with code
└── [your code...]
Aria repository (this repo):
Aria/
├── README.md # This document
├── CLAUDE.md # AI project context
├── VERSION # Version info
├── LICENSE # MIT License
├── standards/ # Methodology specs (submodule)
│ ├── core/ # Core definitions (ten-step cycle)
│ ├── openspec/ # Requirement spec format
│ └── conventions/ # Conventions (git commit, etc.)
├── aria/ # Aria Plugin (submodule)
│ ├── skills/ # 41 Skills (34 user-facing + 7 internal)
│ ├── agents/ # 11 Agents (with STCO descriptions + capabilities)
│ └── .claude-plugin/ # Plugin configuration
├── aria-plugin-benchmarks/ # Skill benchmark suite
│ ├── ab-suite/ # AB test fixtures
│ └── ab-results/ # AB test result archives
├── docs/ # Research documentation
│ ├── architecture/ # System architecture
│ └── requirements/ # PRD + User Stories
├── tests/ # Test files
└── openspec/ # Aria's own OpenSpec changes
└── archive/ # Completed change archives
Project Status
Project Version: 1.7.0
Plugin Version: 1.46.5 (aria-plugin, 41 Skills + 11 Agents)
Maturity: Core workflows verified + project adaptation
PRD v2.0: Approved (AI autonomous development)
Research Focus: Reproducibility of AI collaboration patterns
Contributing
Contributions and discussions are welcome!
- Fork this repository
- Create your branch (
git checkout -b feature/your-feature) - Follow the ten-step cycle workflow
- Submit a Pull Request
License
MIT License — see LICENSE
Contact
- Repository: https://github.com/10CG/Aria
- Plugin: https://github.com/10CG/aria-plugin
- Standards: https://github.com/10CG/aria-standards
- Email: help@10cg.pub
- Maintainer: 10CG Lab