Foundation
Core Concepts
Understanding these three building blocks will help you use Claude Code more effectively.
Agents
Autonomous actors in fresh isolated context with custom tools, permissions, model, memory, and persistent identity
Use for: Use for parallel work, isolated experiments, or long-running tasks
Commands
Knowledge injected into existing context - simple user-invoked prompt templates for workflow orchestration
Use for: Use for quick workflows, repeating patterns, or team-standardized tasks
Skills
Configurable, preloadable, auto-discoverable workflows with context forking and progressive disclosure
Use for: Use for complex multi-step processes, team playbooks, or reusable patterns