convez.ai
SkillAutomaçãoIntermediário
myrgic

CogOS Skills — Definições Portáteis de Skills para Agentes IA

myrgic/plugins

22 skills portáteis para orquestração, pesquisa e DevOps em agentes IA

Ver no GitHub
BomPluginScore · de 0 a 100

Coleção de 22 skills estruturados em formato SKILL.md compatível com Claude Code, Hermes e outros agentes IA. Cobre orquestração de workflows, pesquisa estruturada, ferramentas de desenvolvimento, validação arquitetural e gerenciamento organizacional. Funciona sem modificação entre múltiplas plataformas via padrão aberto Agent Skills.

Instalação

$npx convez-ai myrgic/plugins

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
100%
Último commit
há 1 semana
Versão
—
Issues abertas
7
Contribuidores
1

Casos de uso

  • Orquestração de workflows multi-fase com planejamento paralelo e deliberação
  • Pesquisa estruturada, design de experimentos e validação científica
  • Triagem de PRs/issues e governança em repositórios GitHub
  • Análise arquitetural, forensics de código e qualidade técnica
  • Integração cross-agent via padrão Agent Skills aberto

Veredito

Solução madura e bem-estruturada para automação de workflows cognitivos em agentes IA. A portabilidade cross-platform via SKILL.md é um diferencial técnico sólido; limite-se a projetos que usam Claude Code ou Hermes para máximo aproveitamento.

README

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

CogOS Skills

Portable skill definitions for Claude Code, Hermes, and compatible AI agents.

Each skill is a SKILL.md file — a structured prompt that gives an agent specialized knowledge for a specific domain. The format follows the Agent Skills open standard and works across Claude Code, Cursor, VS Code Copilot, Gemini CLI, Hermes, and other compatible agents.

Setup

Claude Code (plugin marketplace)

Install the whole collection as a marketplace tap:

claude plugin add myrgic/plugins

Or install a specific plugin package:

claude plugin add myrgic/plugins/cogos-workflow
claude plugin add myrgic/plugins/cogos-substrate
claude plugin add myrgic/plugins/cogos-research

Skills are then available in any Claude Code session. Claude invokes them automatically when the conversation context matches, or you can trigger them explicitly (e.g. /plan-phases, /orchestrate).

To install a single skill manually without the plugin system:

mkdir -p .claude/skills/plan-phases
cp plugins/cogos-workflow/skills/plan-phases/SKILL.md .claude/skills/plan-phases/SKILL.md

Hermes (skills tap)

Add the repo as a Hermes skill tap:

hermes skills tap add myrgic/plugins

Then install individual skills:

hermes skills install plan-phases
hermes skills install orchestrate
hermes skills install cogos-workspace

Or search across all taps:

hermes skills search orchestration
hermes skills search substrate

Skills installed via tap are available in all Hermes sessions and appear in the skills_list() tool output. The same SKILL.md files work in both agents — the format is identical.

Available Skills

PluginSkillsDescription
cogos-workflowplan-phases, execute-plan, critical-review, dispatch-agent, council, cold-start, retrospective, local-review, orchestrate, kanban-closed-loop-supervisorPhased planning, parallel execution, deliberation, closed-loop orchestration
cogos-substratecogos-workspace, substrate-archaeology, structural-discourseCogOS/Myrgic ecosystem orientation, prior-work archaeology, discourse conventions
cogos-researchliterature-research, experiment-design, interdisciplinary-research, physics-validation, lab-engineering, inference-cascadeStructured research methodology
cogos-voicevoiceVoice modality via Mod³ on Apple Silicon
cogos-dev-toolsgit-forensics, technical-writing, code-quality, systems-architectureDeveloper tooling and code quality
cogos-architecturecorpus-cross-referenceCogOS architecture corpus hygiene — cross-check proposed RFCs/ADRs against the substrate corpus
myrgic-orgpr-triage, issue-triageOrg-management skills for the myrgic GitHub org

Total: 22 skills across 7 plugin packages.

Cross-Compatibility

Skills in this repo are written to work in both Claude Code and Hermes without modification:

  • Claude Code loads skills from .claude/skills/*/SKILL.md automatically, or via the plugin marketplace (claude plugin add).
  • Hermes loads skills from ~/.hermes/skills/ and from tap-registered GitHub repos (hermes skills tap add).
  • Other agents (Cursor, VS Code Copilot, Gemini CLI) load skills from .claude/skills/ following the same Agent Skills standard.

The SKILL.md frontmatter is the same format for all agents. Agent-specific metadata (Hermes toolset requirements, Claude Code allowed-tools) lives under metadata.hermes or as standard frontmatter fields and is ignored by agents that don't understand it.

Some skills include a canonical_source frontmatter field pointing to the authoritative copy in a running Hermes or cog workspace instance. In those cases the marketplace copy is a projection — update the canonical source first, then project here.

Skill Bundles (Hermes)

Hermes supports composing multiple skills into a named bundle that loads them together with a shared instruction:

# ~/.hermes/skill-bundles/cogos-operator.yaml
name: cogos-operator
description: Full CogOS operator context — workspace orientation, substrate archaeology, architecture hygiene
skills:
  - cogos-workspace
  - substrate-archaeology
  - corpus-cross-reference
instruction: |
  You are operating in the Myrgic/CogOS substrate. Read existing maps and prior work before
  designing or writing. Update canonical sources first; project to marketplace after.

Activate with /cogos-operator in any Hermes session.

Structure

plugins/
├── cogos-workflow/skills/       # Planning + execution + review + orchestration
├── cogos-substrate/skills/      # Workspace orientation + substrate archaeology
├── cogos-research/skills/       # Research methodology
├── cogos-voice/skills/          # Voice modality (Mod³)
├── cogos-dev-tools/skills/      # Developer tooling
├── cogos-architecture/skills/   # Architecture corpus hygiene
└── myrgic-org/skills/           # Org management

Format

Each skill follows the Agent Skills open standard:

---
name: skill-name
description: "One-line description — what the skill does and when to load it."
version: 1.0.0
platforms: [linux, macos, windows]
metadata:
  hermes:
    tags: [tag1, tag2]
    related_skills: [other-skill]
---

# Skill Name

Body: when to use, method, steps, examples.

Contributing

Skills follow the same PR workflow as the rest of the myrgic org — branch on the upstream repo, squash-merge to main. See CONTRIBUTING.md for details.

If a skill has a canonical source in a running workspace (indicated by canonical_source in frontmatter), update that first, then project here. Don't edit the marketplace copy directly.

License

MIT

Ver o original no GitHub

Bom para

Relacionados