## 1. Installation and Setup ### Authentication paths (API key, Bedrock, Vertex) ### Settings hierarchy (user, project, local) ## 2. Operational Modes ### Build mode (default, action-capable) ### Plan mode (read-only analysis and proposal) ## 3. Sub-Agents ### Definition (Markdown + YAML frontmatter) ### Storage locations (`.claude/agents/`, `~/.claude/agents/`) ### Isolation property (own context window) ### Per-agent model assignment ## 4. Skills ### `SKILL.md` files with frontmatter description ### Discovery via description matching ### Same context window as the conversation (no isolation) ### Unification with slash commands (2026) ## 5. Hooks ### 25 lifecycle event points ### Key events: UserPromptSubmit, PreToolUse, PostToolUse, Stop ### Deterministic execution (no model in the loop) ### Exit codes drive blocking semantics (exit 2 → block) ## 6. Model Context Protocol (MCP) Integration ### `.claude/mcp.json` configuration ### Stdio and HTTP transports ### `/mcp` slash command for inspection ## 7. Permissions System ### Allow / deny lists for Bash and other tools ### Granular matchers (`Bash(npm test:*)`) ## 8. Headless mode ### `claude -p` non-interactive invocation ### `--output-format json` for machine-readable output ### `--max-turns` cap ### References - https://code.claude.com/docs/en/sub-agents - https://docs.anthropic.com/en/docs/claude-code - https://ofox.ai/blog/claude-code-hooks-subagents-skills-complete-guide-2026/ - https://alexop.dev/posts/understanding-claude-code-full-stack/