convez.ai
Skill
atravislewis

claude-skills-structure-guide

atravislewis/claude-skills-structure-guide

Ver no GitHub
InicialPluginScore · de 0 a 100

Instalação

$npx convez-ai atravislewis/claude-skills-structure-guide

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
25%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

README

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

Claude Building Patterns Structure Guide

This is a visual study repository for Anthropic's Claude Code and Agent Skills building patterns.

It is intentionally organized as examples you can inspect in VS Code. The demo Claude files live under examples/, not at this repository root, so opening this guide repo does not accidentally activate demo settings, hooks, or skills.

Start Here

  1. Read STRUCTURE-MAP.md for the whole file map.
  2. Read ANTHROPIC-BUILDING-PATTERNS.md for the official-pattern summary.
  3. Read OFFICIAL-RESEARCH-NOTES.md for the source-backed details that are easy to miss.
  4. Open examples/machine-scope/ to see how managed, user-level, and project-level locations relate.
  5. Open examples/project-scope/ and inspect the .claude/ directory.
  6. Compare CLAUDE.md, .claude/rules/, .claude/skills/, .claude/settings.json, .mcp.json, and .claude/agents/.
  7. Open examples/api-scope/ to see how Agent Skills appear in API requests.
  8. Read concepts/ for short exam-prep notes.

Official Sources Used

Core Mental Model

  • Managed machine locations hold organization policy such as managed settings and managed CLAUDE.md.
  • Machine policy locations hold organization-wide Claude Code policy.
  • User-scope files such as ~/.claude/ apply across many repos on that computer.
  • Project-scope files such as repo/.claude/ and repo/CLAUDE.md travel with one repo.
  • CLAUDE.md: always-loaded guidance and project memory that shapes behavior.
  • .claude/rules/*.md: modular guidance, optionally path-scoped.
  • .claude/skills/<name>/SKILL.md: on-demand workflow instructions with optional support files.
  • .claude/settings.json: enforced client configuration, permissions, hooks, and environment.
  • .mcp.json: project-shared MCP server definitions.
  • .claude/agents/*.md: specialized subagents with their own prompt, tools, and model choices.
  • .claude/hooks/*: scripts called by hook definitions in settings.
  • .claude/commands/*.md: older custom command location; skills are now preferred.
  • API container.skills: Agent Skills in the Messages API, paired with code execution.

What This Is Not

This guide only shows Anthropic study fixtures for Claude Code and Agent Skills.

Ver o original no GitHub

Bom para

Relacionados