メニュー
Executa tarefas de implementação em paralelo usando git worktrees isolados. Adaptado do claude-orchestrator para execução multiplataforma no SDLC Agêntico.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | parallel-workers |
| version | 1.0.0 |
| description | Executa tarefas de implementação em paralelo usando git worktrees isolados. Adaptado do claude-orchestrator para execução multiplataforma no SDLC Agêntico. |
| category | implementation |
| phase | 5 |
| complexity_levels | [2,3] |
| tags | ["parallelization","worktrees","automation","phase-5"] |
| scripts | {"worker_manager":"scripts/worker_manager.py","worktree_manager":"scripts/worktree_manager.sh","state_tracker":"scripts/state_tracker.py"} |
| dependencies | {"system":["git >= 2.25","python >= 3.11","jq >= 1.6"],"python":["pyyaml","gitpython"],"skills":["memory-manager","gate-evaluator"]} |
| hooks | {"pre_spawn":"validate_git_status","post_merge":"cleanup_worktree"} |
| observability | {"loki_labels":[{"skill":"parallel-workers"},{"phase":5},{"worker_id":"dynamic"}],"metrics":["worker_count","task_completion_rate","parallel_efficiency"]} |
| security | {"secrets_isolation":true,"sanitize_env":true,"validate_before_merge":true} |
| integration | {"agents":["code-author","test-author","iac-engineer"],"gates":["phase-5-to-6.yml","security-gate.yml"]} |
| issue | #35 |
| epic | #33 |
Skill para execução paralela de tarefas de implementação usando git worktrees isolados.
Esta skill permite que múltiplas tarefas de código sejam desenvolvidas simultaneamente por workers independentes, cada um operando em seu próprio git worktree. Inspirada no claude-orchestrator, foi adaptada para ser multiplataforma e integrar com o ecossistema do SDLC Agêntico.
┌─────────────────────────────────────────────────────────┐
│ Main Repository │
│ (feature/epic-33-*) │
└────────────┬────────────────────────────────────────────┘
│
├─→ Worker 1: ~/.worktrees/project/task-001/
│ Branch: feature/task-001
│ State: WORKING
│ Agent: code-author
│
├─→ Worker 2: ~/.worktrees/project/task-002/
│ Branch: feature/task-002
│ State: PR_OPEN
│ Agent: test-author
│
└─→ Worker 3: ~/.worktrees/project/task-003/
Branch: feature/task-003
State: NEEDS_INIT
Agent: iac-engineer
UNKNOWN ──→ NEEDS_INIT ──→ WORKING ──→ PR_OPEN ──→ MERGED
↑ ↑ ↓ ↓
└────────────┴─────────────┴────────────┘
(error recovery)
States:
UNKNOWN: Initial or error stateNEEDS_INIT: Worker created, needs initializationWORKING: Task in progressPR_OPEN: Pull request created, awaiting reviewMERGED: Task completed and merged# Spawn a single worker
python3 .claude/skills/parallel-workers/scripts/worker_manager.py spawn \
--task-id "TASK-001" \
--description "Implement user authentication" \
--agent "code-author" \
--base-branch "feature/epic-33-claude-orchestrator"
# Spawn multiple workers from spec
python3 .claude/skills/parallel-workers/scripts/worker_manager.py spawn-batch \
--spec-file .agentic_sdlc/projects/current/tasks.yml
# Create worktree
.claude/skills/parallel-workers/scripts/worktree_manager.sh create \
project-name task-001 feature/epic-33
# List active worktrees
.claude/skills/parallel-workers/scripts/worktree_manager.sh list project-name
# Remove worktree
.claude/skills/parallel-workers/scripts/worktree_manager.sh remove \
project-name task-001
# Check worker state
python3 .claude/skills/parallel-workers/scripts/state_tracker.py get worker-001
# List all workers
python3 .claude/skills/parallel-workers/scripts/state_tracker.py list
# Update state
python3 .claude/skills/parallel-workers/scripts/state_tracker.py set \
worker-001 WORKING
Workers operate with sanitized environments:
# Sanitized variables (kept)
- PATH
- HOME
- USER
- LANG
# Removed from worker env
- *_KEY
- *_SECRET
- *_TOKEN
- *_PASSWORD
All workers must pass quality gates before merge:
# From security-gate.yml
- no_hardcoded_secrets
- input_validation_present
- sast_scan_passed
All worker operations logged to Loki:
{
"skill": "parallel-workers",
"phase": 5,
"worker_id": "worker-001",
"task_id": "TASK-001",
"state": "WORKING",
"timestamp": "2026-01-21T10:30:00Z",
"correlation_id": "abc-123-def"
}
# delivery-planner generates tasks
tasks = [
{"id": "TASK-001", "agent": "code-author", "desc": "..."},
{"id": "TASK-002", "agent": "test-author", "desc": "..."},
]
# Spawn workers in parallel
for task in tasks:
spawn_worker(task)
# Wait for all workers to complete
all_merged = wait_for_workers(timeout=3600)
if all_merged:
# Run phase-5-to-6 gate
gate_result = evaluate_gate("phase-5-to-6.yml")
if gate_result.passed:
transition_to_phase(6)
# All parallel-workers activity
{skill="parallel-workers"}
# Errors by worker
{skill="parallel-workers", level="error"} | json | line_format "{{.worker_id}}: {{.message}}"
# Task completion timeline
{skill="parallel-workers", state="MERGED"} | json | line_format "{{.task_id}} completed at {{.timestamp}}"
New panels added to .claude/config/logging/dashboards/sdlc-overview.json:
| Metric | Sequential | Parallel (3 workers) | Improvement |
|---|---|---|---|
| Total time (3 tasks) | 90 min | ~35 min | 61% faster |
| Resource usage | 1x CPU | 3x CPU | Scalable |
| Context switching | None | Minimal | Isolated |
| Merge conflicts | N/A | Zero (worktrees) | Safe |
.project/corpus/nodes/learnings/LEARN-claude-orchestrator-patterns.ymlOrquestrador central do SDLC Agêntico. Coordena todas as 8 fases do ciclo de desenvolvimento, gerencia transições entre fases, aplica quality gates, mantém contexto persistente e garante que todas as fases foram executadas corretamente. Use este agente para: - Iniciar novos workflows SDLC - Gerenciar transições entre fases - Avaliar quality gates - Escalar decisões para humanos - Coordenar múltiplos agentes Examples: - <example> Context: Usuário quer iniciar um novo projeto user: "Preciso desenvolver uma API de pagamentos" assistant: "Vou usar @orchestrator para iniciar o workflow SDLC completo" <commentary> Novo projeto requer todas as fases do SDLC, começando pela Fase 0 (Intake) </commentary> </example> - <example> Context: Fase atual completou, precisa avançar user: "Os requisitos estão prontos, podemos avançar?" assistant: "Vou usar @orchestrator para avaliar o gate da Fase 2 e decidir se podemos avançar para Arquitetura" <commentary> Transição entre fases requer avaliação de ga
Integracao com BMAD Method para escala adaptativa e workflows guiados. Detecta nivel de complexidade e ajusta agentes automaticamente. Use quando: iniciar workflow, detectar complexidade, mapear agentes BMAD.
Sistema de scoring temporal para nodes de conhecimento do corpus RAG. Calcula scores de obsolescencia baseado em idade, validacao e acesso. Gera sugestoes de curadoria para conteudo desatualizado.
Avalia quality gates entre fases do SDLC. Verifica artefatos obrigatorios, criterios de qualidade, e aprovacoes necessarias antes de permitir transicao. Use quando: transicao entre fases, verificacao manual de gate, auditoria de qualidade.
Gerencia persistencia de contexto, decisoes e learnings do projeto. Armazena e recupera informacoes entre sessoes para manter continuidade. Use quando: salvar decisoes, recuperar contexto, persistir learnings.
Gerencia commits automaticos ao final de cada fase do SDLC. Garante que artefatos de cada fase sejam commitados e rastreados. Use quando: transicao de fase, checkpoint de progresso.