convez.ai
SkillAutomaçãoAvançado
lewishowles

Agents Configuration Global

lewishowles/agents

Config centralizada de regras, skills e hooks para Claude, Codex e ChatGPT

Ver no GitHub
InicialPluginScore · de 0 a 100

Repositório de configuração centralizada compartilhada para Claude Code, OpenAI Codex e ChatGPT. Mantém regras comuns, skills, hooks e templates de projetos em um único lugar, com scripts para gerar arquivos específicos de cada ferramenta e facilitar setup global e por-projeto.

Instalação

$npx convez-ai lewishowles/agents

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
0
Forks
0
Acompanhando
0
Saúde
28%
Último commit
há 2 dias
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Padronizar regras e comportamentos entre múltiplos agentes IA
  • Gerenciar skills reutilizáveis de forma versionada e sincronizada
  • Automatizar setup de Claude Code e OpenAI Codex em múltiplos projetos
  • Manter biblioteca de templates e hooks de projeto consistente

Veredito

Ferramenta robusta e bem estruturada para quem trabalha com múltiplos agentes IA e precisa de configuração padronizada em escala. Shell scripts sólidos, mas requer familiaridade com ambiente Unix/Linux e setup manual.

README

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

Global agent configuration

Shared configuration for Claude Code and OpenAI Codex, and ChatGPT.

This repository keeps common agent rules, skills, hooks, and project templates in one place. The setup is based on how I work as a designer and developer: small reviewable changes, accessible interfaces, maintainable code, clear communication, and tools that make repeated work easier.

The included scripts generate the target files each tool expects for global and per-project setup.

What's inside

  • rules/ - source fragments used by both Claude and Codex
  • dist/claude/ - generated CLAUDE.md, Claude settings, hooks, and Claude-only source fragments
  • dist/codex/ - generated AGENTS.md and Codex-only source fragments
  • dist/chatgpt/ - generated SKILLS.md index and per-skill files for upload to a ChatGPT Custom GPT knowledge base
  • skills/ - user skills, either flat (skills/<name>/) or grouped (skills/<group>/<name>/)
  • external-skills.json - official upstream skills synced into skills/
  • scripts/ - sync and setup scripts
  • templates/ - project templates for Claude, Codex, or both
  • docs/ - deeper reference: setup, Codex, hooks, skills, commands, agents, plugins

Initial setup

Replace /path/to/repository with the path to this repository.

cd /path/to/repository
scripts/setup-global.sh --both

Use --claude or --codex to configure one runtime only. With no flag, the script asks which agent(s) to configure.

The global setup script syncs official external skills, runs scripts/sync.sh, then links:

  • ~/.claude/CLAUDE.md to dist/claude/CLAUDE.md
  • ~/.claude/settings.json to dist/claude/settings.json
  • ~/.claude/.mcp.json to dist/claude/.mcp.json
  • ~/.claude/skills/<name> to skills/<name> or skills/<group>/<name>
  • ~/.claude/hooks/<file> to dist/claude/hooks/<file>
  • ~/.agents/AGENTS.md to dist/codex/AGENTS.md
  • ~/.codex/AGENTS.md to dist/codex/AGENTS.md
  • ~/.agents/skills/<name> to skills/<name> or skills/<group>/<name>
  • ~/.codex/skills/<name> to skills/<name> or skills/<group>/<name>

It also ensures ~/.codex/config.toml has the codebase-memory-mcp MCP server entry.

Existing files are backed up instead of overwritten.

If you need to run setup without network access, use:

scripts/setup-global.sh --both --skip-external

ChatGPT setup

ChatGPT doesn't support automatic skill loading, so the setup is manual. After running scripts/sync.sh, the dist/chatgpt/ directory contains everything you need.

System prompt — paste the contents of dist/chatgpt/INSTRUCTIONS.md into your Custom GPT's system prompt (or into ChatGPT's custom instructions).

Knowledge base — upload all the other files in dist/chatgpt/ to the Custom GPT's knowledge base: SKILLS.md plus one .md file per skill.

Once uploaded, you can reference skills explicitly or let ChatGPT retrieve them automatically:

  • "Use my vue and code-style skills." — explicit, most reliable
  • "Use my skills." — ChatGPT reads SKILLS.md, determines which are relevant, and loads them

Re-upload the files after running scripts/sync.sh whenever skills change.

Project setup

From a project root:

/path/to/repository/scripts/setup-project.sh --both

Use --claude, --codex, or --both:

  • --claude creates AGENTS.md, .claude/settings.json, .claude/.claudeignore, and .claude/templates/PLAN.md.template
  • --codex creates AGENTS.md
  • --both creates shared AGENTS.md and the Claude .claude/ files

Project setup skips existing files. It does not overwrite or back up project files because those are likely hand-edited.

Hook dependency

Claude skill-trigger hooks require jq:

brew install jq

Codex hooks are separate from the Claude hooks in this repo. See docs/codex.md for the current Codex behaviour.

Shell aliases

Add aliases to ~/.zshrc if you run setup often:

alias setup:agents:global="/path/to/repository/scripts/setup-global.sh --both"
alias setup:claude:global="/path/to/repository/scripts/setup-global.sh --claude"
alias setup:codex:global="/path/to/repository/scripts/setup-global.sh --codex"
alias setup:agents="/path/to/repository/scripts/setup-project.sh --both"
alias setup:claude="/path/to/repository/scripts/setup-project.sh --claude"
alias setup:codex="/path/to/repository/scripts/setup-project.sh --codex"

Common commands

scripts/sync.sh
scripts/sync-external-skills.sh
scripts/setup-global.sh --both
scripts/setup-project.sh --both
tests/setup-project.sh

Going deeper

Ver o original no GitHub

Bom para

Relacionados