Stats
Actions
Tags
From dev-team-kit-fv
Routes tasks to the appropriate declarative program (programs/*.yml) based on task classification. Works with intent-classifier hook and orchestrator skill. Use when a task maps to an existing program rather than improvising a pipeline.
How this skill is triggered — by the user, by Claude, or both
Slash command
/dev-team-kit-fv:39-program-routerThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Decide qual program declarativo rodar baseado em classificação da task. Trabalha em par com:
Decide qual program declarativo rodar baseado em classificação da task. Trabalha em par com:
intent-classifier (auto-sugere antes do agente começar)Esta skill segue GLOBAL.md, policies/auto-orchestration.md, policies/handoffs.md, policies/execution.md e policies/programs-schema.md.
/run-program <nome> explícito → rodar exatamente isso. Skill não interfere.intent-classifier sugeriu program e usuário concordou → essa skill confirma e dispatcha./run-program explícitomemory/constitution.md — pode forçar pipeline específicorouted: program + inputs sugeridosdecline: razão + handoff para skill 09 ou ação direta| Program | Use case | Confidence sinais |
|---|---|---|
pipeline-discovery | Ideia vaga → discovery formal → issues | "ideia vaga", "preciso de PRD", "grill-me" |
spec-driven-development | Feature em projeto maduro com constitution | "nova feature", "constitution", "spec-driven" |
loop-polishing | Task autônoma com polish pré-commit | "auto-loop", "autônomo", "fire and forget" |
detective-spec | Reverse-engineering de legado | "legacy", "sem docs", "extrair contratos" |
adversarial-dev | App from-scratch com GAN-style adversarial loop | "from scratch", "greenfield", "construir app" |
comprehensive-review | PR review profundo (5 agents) | "review crítico", "5-agent", "comprehensive review" |
additionalContext do hook intent-classifier (geralmente sim, então a decisão é confirmar)Match heurístico em ordem de prioridade:
memory/constitution.md. Se declara "todo feature passa por spec-driven-development", forçar.adversarial-devspec-driven-developmentpipeline-discoverycomprehensive-reviewdetective-specloop-polishingSe nenhum match: decline → handoff skill 09.
Vou rotear esta task para o program `<nome>` porque <razão concreta>.
Opções:
- ✅ Rodar dry-run primeiro (recomendado) — mostra plano antes de executar
- ▶️ Rodar direto (gates humanos pausam mid-flow)
- 🛠️ Pipeline ad-hoc (devolve para skill 09 montar fluxo custom)
- ❌ Cancelar
Se "dry-run":
node scripts/run-program.mjs <program-id> --dry-run --input key=value
Apresentar plano resolvido. Esperar nova confirmação.
Se "direto":
Invocar /run-program <program-id> com inputs colhidos.
Se "ad-hoc": Handoff para skill 09 (orchestrator). Skill 09 monta pipeline informal.
Se "cancelar": Devolver controle para conversa normal.
/run-program <name> — caminho default quando match high confidenceListar programs disponíveis sem rotear:
node scripts/run-program.mjs --list
Inspecionar program específico antes de propor:
node scripts/run-program.mjs --describe <program-id>
npx claudepluginhub felvieira/claude-skills-fv --plugin dev-team-kit-fvOrchestrates development pipeline by classifying tasks, coordinating skill execution order, adapting flow to context, and ensuring no critical steps are skipped.
Routes every agent action through correct skills via routing table before code, files, or technical decisions. Checks adaptive overrides; defaults to rune:cook for code tasks.