jeffecom

Sistema de workflow unificado para Claude Code — pipeline de orquestracao, memory bank, hooks e quality gates.

Instalacao

npx jeffecom

Isso instala no diretorio do seu projeto:

• .claude/hooks/ — 6 hooks + 5 sub-modulos (_lib facade + parse, config, memory, hooks, vault)
• .claude/commands/ — 21 slash commands (14 essenciais + 5 aliases + wrappers)
• .claude/skills/ — 56 skills (35 infra, 6 level, 5 docs, 4 standalone, 4 utility, 2 meta)
• .claude/rules/ — 5 regras de qualidade + rules decision matrix
• .claude/config/ — configuracao centralizada
• memory-bank/ — documentacao persistente
• CLAUDE.md — instrucoes do projeto

Com BMAD-METHOD (opcional)

npx jeffecom --with-bmad

Adiciona 12+ agent personas, 30+ workflows step-file (PRD, architecture, code review, sprint) do BMAD-METHOD.

Uso

Novo usuario? Comece pelo Guia de Inicio Rapido.

Apos instalar, abra Claude Code no projeto e use:

/configurar          # Configura MCPs, permissoes, plugins, safety
/carregar            # Carrega contexto do projeto (7 fases)
/orq <descricao>     # Executa pipeline de orquestracao (ate 9 fases)

Commands

| Comando | Descricao | |---------|-----------| | /orq | Pipeline de orquestracao (ate 9 fases, decomposto em step files) | | /configurar | Configuracao interativa (MCPs, permissoes, plugins, safety) | | /validate | Validacao completa da stack (type-check, lint, testes, build) | | /status | Status do projeto (git, memory bank, saude) | | /carregar | Carrega contexto do projeto (7 fases) | | /compact | Compacta contexto (salva checkpoint antes) | | /metrics | Metricas da sessao de trabalho | | /release | Release com semantic versioning | | /restore | Restaura checkpoint do memory bank | | /code-review | Code review adversarial (→ skill) | | /security-audit | Auditoria de seguranca OWASP (→ skill) | | /analyze | Analise de escopo e complexidade (→ skill) | | /refactor | Refatoracao standalone com metricas (→ skill) | | /deploy | Deploy com multiplas estrategias (→ skill) | | /docs | Gera/atualiza documentacao (→ skills docs) | | /prd | Criar PRD guiado (→ skill prd-generator) |

Aliases deprecados: /bootstrap→/configurar, /setup→/configurar, /squad→/orq, /agents→/orq, /reviewer→/code-review

Skills (56)

| Categoria | Qtd | Exemplos | |-----------|-----|----------| | Infra: Monitoring | 8 | status-check, log-viewer, metrics-collector, uptime-monitor | | Infra: CI/CD | 8 | ci-status, ci-trigger, env-manage, rollback, preview-deploy | | Infra: Advanced | 8 | db-migrate, db-backup, scale, cache-clear, performance-check | | Infra: Core | 11 | deploy, restart, logs, health-check, backup, alerts, traffic | | Level | 6 | l0-micro a l5-rewrite (task sizing) | | Docs | 5 | prd-generator, design-doc, runbook, trd-rfc, security-report | | Standalone | 4 | analyzer, code-reviewer, refactorer, security-audit | | Utility | 4 | dependency-auditor, diagram-generator, memory-checkpoint, project-maturity | | Meta | 2 | skill-creator, claude.ai |

Pipeline de Orquestracao

FASE 1: ANALYZE   → Classifica L0-L5, mapeia escopo e riscos
FASE 2: PLAN      → Cria plano com steps e criterios (L1+)
FASE 3: IMPLEMENT → Executa plano, edita codigo, valida types
FASE 4: DOCS      → Gera/atualiza documentacao (L2+)
FASE 5: TEST      → Roda tsc, testes, valida criterios (L1+)
FASE 6: SECURITY  → Audit OWASP, deps, secrets (L3+)
FASE 7: REVIEW    → Code review, qualidade, aderencia (L1+)
FASE 8: REFACTOR  → Simplificacao DRY/KISS (L4+)
FASE 9: DEPLOY    → Deploy com rollback (L5 obrigatorio, L2+ condicional)

Cada fase tem seu proprio step file em commands/orq/step-*.md.

Hooks

| Hook | Evento | Funcao | |------|--------|--------| | session-start | SessionStart | Contexto + infra + stack + config + MCP detection + pipeline state | | post-edit-pipeline | PostToolUse (Edit/Write) | Quality checks + lint fix + tracking + changelog nudge | | context-monitor | PostToolUse (Edit/Write) | Warnings 15%/8% (1M ctx) + rules reinforcement + smart compact suggestion | | save-checkpoint | PreCompact | Checkpoint ACTIVE.md + checkpoint retention (age + count) | | update-active | Stop | Lembrete ACTIVE.md + session compaction | | statusline | Notification | Barra visual minima |

Shared library: _lib.cjs (facade, 42 exports) + 5 sub-modulos (_lib-parse, _lib-config, _lib-memory, _lib-hooks, _lib-vault).

Configuracao

Apos instalar, use /configurar para customizar:

• MCPs: Context7, Playwright, Figma, Linear, Notion, Chrome (auto-detectados)
• Permissoes: Permissivo, Moderado ou Restrito
• Pipeline: Interactive ou Yolo
• Quality Gates: Plan approval, testes, review, docs
• Safety: Bloqueia .env, secrets, force push
• TDD: Opcional (red-green-refactor)
• Checkpoint Retention: Por idade (30d) + contagem (20) + minimo (5)

Configs armazenadas em .claude/config/config.json e .claude/settings.json.

BMAD-METHOD

O jeffecom integra opcionalmente o BMAD-METHOD, um framework de desenvolvimento agil com IA.

A integracao adiciona:

• Agent Personas: Agents ganham identidade, estilo e principios
• 30+ Workflows: PRD guiado, code review adversarial, sprint planning, research estruturado
• Step-File Architecture: Micro-files sequenciais com just-in-time loading
• WIP Tracking: Resume automatico de trabalho interrompido

Re-instalacao

npx jeffecom --force    # Sobrescreve hooks, commands, rules
npx jeffecom --dry-run  # Mostra o que seria feito

Requisitos

• Node.js >= 18.0.0
• Claude Code CLI

Licenca

MIT