convez.ai
AgenteDesenvolvimento WebAvançado
Emasoft

AI Maestro Architect Agent

Emasoft/ai-maestro-architect-agent

Converte requisitos em documentos de arquitetura e especificações técnicas

Ver no GitHub
InicialPluginScore · de 0 a 100

Agent especializado em transformar requisitos em documentos de design revisados e implementáveis. Realiza análise de requisitos, pesquisa de APIs, decisões arquiteturais e decomposição modular, criando especificações técnicas que coordenam trabalho de implementação.

Instalação

$npx convez-ai Emasoft/ai-maestro-architect-agent

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
2
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 1 mês
Versão
v2.17.21
Issues abertas
1
Contribuidores
1

Casos de uso

  • Análise e documentação de requisitos de projeto
  • Design de arquitetura e especificações técnicas
  • Pesquisa e planejamento de integrações de API
  • Decomposição modular e planejamento de implementação
  • Design de pipelines CI/CD e workflows

Veredito

Ferramenta robusta para equipes que precisam sistematizar a fase de design antes da implementação. Integra-se bem em ecossistemas multi-agent onde coordenação entre análise, arquitetura e execução é crítica.

README

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

Architect Agent (amaa-)

Claude Code role-plugin for the AI Maestro multi-agent ecosystem: the ARCHITECT (AMAA) persona that turns requirements into reviewed, implementable design documents — requirements analysis, API research, architecture decisions, module breakdowns, and handoff packages.

Version: 2.7.0

Overview

The Architect Agent handles design documents, requirements analysis, and architecture decisions. It creates specifications that the Orchestrator uses to coordinate implementation work.

Prefix: amaa- = AI Maestro Architect Agent

Core Responsibilities

  1. Requirements Analysis: Gather and document requirements
  2. Design Documents: Create technical specifications and architecture docs
  3. API Research: Investigate APIs and integration points
  4. Planning: Break down work into implementable modules
  5. Hypothesis Verification: Test assumptions before committing to design

Components

Agents

AgentDescription
ai-maestro-architect-agent-main-agent.mdMain architect agent
amaa-documentation-writer.mdCreates technical documentation
amaa-api-researcher.mdResearches APIs and integrations
amaa-modularizer-expert.mdBreaks work into modules
amaa-planner.mdCreates implementation plans
amaa-cicd-designer.mdDesigns CI/CD pipelines and workflows

Commands

CommandDescription
amaa-start-planningStart planning phase
amaa-add-requirementAdd new requirement
amaa-modify-requirementModify existing requirement
amaa-remove-requirementRemove requirement

Skills

SkillDescription
amaa-design-lifecycleDesign document management
amaa-requirements-analysisRequirements patterns
amaa-documentation-writingDocumentation skills
amaa-api-researchAPI research patterns
amaa-planning-patternsPlanning methodology
amaa-hypothesis-verificationTest assumptions
amaa-design-communication-patternsShared utilities
amaa-cicd-designCI/CD pipeline design patterns
amaa-design-managementDesign document management tools
amaa-github-integrationGitHub integration patterns
amaa-label-taxonomyLabel and tagging patterns
amaa-modularizationModule decomposition patterns
amaa-session-memorySession context persistence
architect-memory-recallSymptom-ranked recall over markdown memory notes (memgrep with grep fallback)
architect-memory-writeCapture one durable, symptom-indexed memory note + index line

Hooks

HookEventDescription
amaa-stop-checkStopBlock exit until all design work is complete

Workflow

  1. Receives requirements from AMCOS (Chief of Staff — the team boundary)
  2. Analyzes requirements and creates design documents
  3. Breaks work into implementable modules
  4. Creates handoff document for AMOA (Orchestrator)
  5. Reports completion to AMCOS (Chief of Staff)

Per the R6 v3 communication graph, the ARCHITECT is a team-internal role: work intake and completion reporting flow through AMCOS (the Chief of Staff guards the team boundary); the MANAGER (AMAMA) reaches the team only via AMCOS. A direct AMAA→AMOA edge exists for design handoffs.

Output Artifacts

  • Design documents (markdown)
  • Module specifications
  • API integration plans
  • Architecture diagrams (mermaid)
  • Handoff files for AMOA (Orchestrator)

Recommended MCP Servers

The architect agent benefits from these MCP servers when available:

MCP ServerPurpose
LLM Externalizer (llm-externalizer)Offload file analysis/scanning to external LLMs, saving orchestrator context tokens
Serena MCP (serena-mcp)Symbol-level code navigation (find definitions, references, overviews)

The tldr CLI tool is also recommended for token-efficient code structure analysis (tldr structure, tldr arch, tldr imports).

Installation

From Git Repository (recommended)

Install directly from the GitHub repo using the git-subdir source type (points to a subdirectory within the repo):

claude plugin install --source git-subdir --url https://github.com/Emasoft/ai-maestro-architect-agent --scope local

After installing, activate changes without restarting:

/reload-plugins

Then start a session with the main agent:

claude --agent ai-maestro-architect-agent-main-agent

Development Only (--plugin-dir)

--plugin-dir loads a plugin directly from a local directory without installation. Use only during plugin development.

claude --plugin-dir ./ai-maestro-architect-agent

Usage

Start a session as the architect:

claude --agent ai-maestro-architect-agent-main-agent

Then drive the planning workflow with the slash commands:

/amaa-start-planning Build a REST API for inventory management
/amaa-add-requirement REQ "User authentication" --criteria "JWT-based login" --priority high
/amaa-modify-requirement REQ R-001 --status approved
/amaa-remove-requirement REQ R-002

The agent produces design documents under docs_dev/design/, registers each one in the design index with a UUID, walks it through the DRAFT → REVIEW → APPROVED → IMPLEMENTING → COMPLETED → ARCHIVED lifecycle, and prepares handoff packages for the Orchestrator (AMOA). The Stop hook blocks session exit while draft designs, pending tasks, orphan requirements, or open architect-assigned GitHub issues remain (capped at 3 blocks per session via CLAUDE_CODE_STOP_HOOK_BLOCK_CAP).

Non-Standard Directories

DirectoryPurpose
lib/Shared constants, templates, and schemas used across skills and agents

The pre-push gate is installed automatically by scripts/publish.py (it writes .githooks/pre-push and activates core.hooksPath); plugin validation runs only via the remote CPV launcher (uvx cpv-remote-validate), not vendored locally.

Platform Requirements

  • Sync script: Cross-platform Python script. Requires gh CLI authenticated.
  • Python scripts: Cross-platform. Use uv run --with pyyaml python <script> or python3 <script>.

Validation

Whole-plugin validation uses the remote CPV validator (always current, never vendored):

uvx cpv-remote-validate plugin . --strict
Ver o original no GitHub

Bom para

Relacionados