convez.ai
SkillAutomaçãoAvançado
dremonkey

Agents and Skills para Claude Code

dremonkey/agents-and-skills

Suite de skills para planejamento, revisão técnica e orquestração de projetos em Claude Code

Ver no GitHub
InicialPluginScore · de 0 a 100

Coleção de definições de agentes e prompts de skills para Claude Code, cobrindo planejamento estratégico, revisão técnica, orquestração de iniciativas e análise de MRs/PRs. Fornece roles especializados (CEO, CTO, Engineering Manager, Architect) para automação de fluxos de desenvolvimento completos.

Instalação

$npx convez-ai dremonkey/agents-and-skills

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 3 meses
Versão
—
Issues abertas
0
Contribuidores
2

Casos de uso

  • Orquestração end-to-end de iniciativas (PRD → arquitetura → tasks)
  • Planejamento de tarefas com dependency-awareness para equipes de engenharia
  • Revisão técnica profunda (arquitetura, riscos, segurança, testes, rollout)
  • Revisão de MRs/PRs com triagem automática (PLAN, IMPL-FIRST, FOLLOWUP)
  • Rascunho de arquiteturas técnicas com system design e tradeoffs
  • Desafio estratégico de planos em modo CEO (expand/hold/reduce)

Veredito

Conjunto robusto de skills Go Template para orquestração complexa de fluxos de desenvolvimento. Ideal para equipes que usam Claude Code/Cursor e precisam automatizar planning, revisão técnica e coordenação estratégica — requer compreensão profunda de workflows de engenharia para máxima efetividade.

README

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

Agents and Skills

A Claude Code plugin packaging agent definitions and skill prompts.

Skills

Skill nameRoleShort description
plan-eng-tasksEngineering manager execution leadConverts approved scope into dependency-aware task plans and drives execution via structured tasks and sub-agents.
plan-requirement-docCEO/founder strategic reviewerChallenges plans at product and strategic levels with expansion/hold/reduction review modes.
plan-tech-specCTO technical reviewerPerforms deep technical readiness review across architecture, risks, security, testing, and rollout.
orchestrate-initiative-planningPlanning workflow orchestratorCoordinates CEO review, CTO review, and engineering task planning from PRD through EPICs and task files without implementation.
draft-technical-architectureTechnical architectProduces architecture drafts with clear system design, tradeoffs, and implementation direction.
review-mrMR/PR reviewer (GitLab + GitHub)Reviews a GitLab MR (glab) or GitHub PR (gh) — auto-detects the platform via a provider file — in one of three triaged modes: PLAN (planning prose), IMPL-FIRST (first-pass code review), or FOLLOWUP (delta re-review). Publishes a top-level + line-tied review.

Structure

.claude-plugin/
  plugin.json                        — Plugin manifest
setup.sh                             — Symlink helper for Cursor-compatible harnesses
skills/
  <skill-name>/
    SKILL.md

Installation

Claude Code (symlink)

Symlink each skill into ~/.claude/skills/:

for skill in skills/*/; do
  ln -sfn "$(pwd)/$skill" ~/.claude/skills/"$(basename "$skill")"
done

Or use the helper script:

./setup.sh ~/.claude/skills

Skills will be available as slash commands (e.g., /plan-eng-tasks) in your next session.

Claude Code (per-session plugin)

Load the entire repo as a plugin for a single session:

claude --plugin-dir /path/to/agents-and-skills

Cursor (and other compatible harnesses)

Use the symlink helper script to link every skill into your local skills directory:

# Default target: ~/.cursor/skills
./setup.sh

# Optional: provide a custom target directory
./setup.sh "$HOME/.cursor/skills"

The script is idempotent: it keeps existing correct links and only creates or updates missing links.

Dependencies

Some skills reference other skills that must be installed separately:

  • excalidraw-diagram-skill — used by draft-technical-architecture for system-level Excalidraw diagrams

Usage

Skills are invoked via slash commands in Claude Code (e.g., /plan-eng-tasks).

Artifact path conventions

Use these canonical paths for planning and execution artifacts:

  • Product strategy artifacts: vision/<INITIATIVE>/VISION.md and vision/<INITIATIVE>/PRD.<TOPIC>.md
  • Technical architecture artifacts: docs/architecture/<INITIATIVE>/
  • Execution artifacts (epics + tasks): tasks/<EPIC_NAME>/

Recommended handoff flow:

  • plan-requirement-doc -> vision/<INITIATIVE>/...
  • plan-tech-spec -> docs/architecture/<INITIATIVE>/ + tasks/<EPIC_NAME>/EPIC.md
  • plan-eng-tasks -> updates tasks/<EPIC_NAME>/... and docs/architecture/<INITIATIVE>/ as implementation details are discovered
  • orchestrate-initiative-planning -> coordinates the full handoff across all three stages

Authoring guidance

  • Claude-focused guidance: CLAUDE.md
  • Cursor-focused guidance: CURSOR.md

License

MIT

Ver o original no GitHub

Bom para

Relacionados