tools-only Anthropic Best Practices
Consolidated from Anthropic's official documentation and guides (February 2026)
Sources:
docs/The-Complete-Guide-to-Building-Skill-for-Claude.pdf(Anthropic, Feb 2026)- Claude Code Skills Documentation (
code.claude.com/docs/skills)- Claude Code Subagents Documentation (
code.claude.com/docs/sub-agents)- Claude Code Agent Teams Documentation (
code.claude.com/docs/agent-teams)- Claude Code Memory Documentation (
code.claude.com/docs/memory)- Claude Code Hooks Documentation (
code.claude.com/docs/en/hooks)- Directive Compliance Framing research
Document Index
This guide is split into focused reference files. Read only the file relevant to your current task.
| File | Sections | Lines | When to Read |
|---|---|---|---|
| Skills Guide | 1-18 | ~960 | Building, testing, iterating, or distributing Agent Skills |
| Claude Code Agents | 19-21 | ~250 | Configuring Claude Code skill extensions, subagents, or agent teams |
| Hooks Reference | 25 | ~650 | Setting up or debugging lifecycle hooks |
| Memory & Prompting | 22-24 | ~140 | Writing CLAUDE.md rules, directive framing, docs architecture |
| Takeaways | 26 | ~50 | Quick-reference action checklist |
Quick Navigation
Building a Skill?
Start with the Skills Guide:
- Section 5: YAML frontmatter (the #1 factor for triggering)
- Section 6: Writing effective instructions
- Section 12: Workflow patterns (5 reusable templates)
- Section 14: MCP tool usage lessons learned
- Section 16: Pre-release checklist
Configuring Agents?
See Claude Code Agents:
- Section 19: Claude Code-specific skill features (frontmatter fields, dynamic injection)
- Section 20: Subagent best practices (custom agents, permission modes, skill preloading)
- Section 21: Agent teams (experimental, multi-instance coordination)
Setting Up Hooks?
See Hooks Reference:
- All 14 event types with input/output schemas
- Hook handler types (command, prompt, agent)
- Configuration structure with matcher examples
- 5 design patterns from our actual setup
Writing CLAUDE.md or Rules?
See Memory & Prompting:
- Section 22: Memory hierarchy, imports, modular rules
- Section 23: Directive compliance framing (welfare-based > imperative)
- Section 24: LLM-optimized documentation principles
Quick Reference?
See Takeaways:
- 33 actionable insights with source attribution
- Covers skills, agents, hooks, memory, and MCP integration