name: claude-code-best-practices description: Reference for Claude Code features and best practices: writing CLAUDE.md, skills, hooks, sub-agents, agent teams, headless mode, scheduled tasks, worktrees, batch, frontend verification. Use when authoring/auditing CLAUDE.md, configuring Claude Code, writing skills, or orchestrating agents. Keywords CLAUDE.md, skill, hook, sub-agent, agent team, headless, worktree, /loop. allowed-tools: Read, Grep, Glob, WebFetch, WebSearch, Bash(cat ), Bash(gh ), Bash(git ), Bash(ls )

Claude Code Best Practices

Reference knowledge for Claude Code features. Each topic has a local summary and a canonical URL.

Prime Directive: CLAUDE.md Is Load-Bearing

CLAUDE.md is the only file that goes into every Claude Code conversation by default. Every line competes for a finite instruction budget (~150–200 instructions, of which ~50 are already spent on the system prompt). Authoring or editing a CLAUDE.md — global, project, or skill-scoped — is always a high-leverage operation.

Whenever the user asks you to write, refactor, audit, or add to a CLAUDE.md, load writing-claude-md first and apply it. Proactively flag bloat, task-specific rules, linter-style guidance, and long inline examples. Prefer extracting deep content into agent_docs/ (or equivalent) with a one-line pointer from the root file.

Fetch Strategy

Quick answers: use the local summary in the reference file
Before making changes or giving definitive guidance: WebFetch the canonical URL to get the latest docs, then act on the live content

Topic Routing

Topic	Reference	Canonical URL
Writing a good CLAUDE.md	writing-claude-md	https://www.humanlayer.dev/blog/writing-a-good-claude-md
Built-in tool catalog & behavior	tools-reference	https://code.claude.com/docs/en/tools-reference
Creating and configuring skills	skills	https://code.claude.com/docs/en/skills
Skill authoring best practices	skill-best-practices	https://platform.claude.com/docs/en/agents-and-tools/agent-skills/best-practices
Spawning isolated subagents	sub-agents	https://code.claude.com/docs/en/sub-agents
Multi-session agent teams	agent-teams	https://code.claude.com/docs/en/agent-teams
Lifecycle hooks and automation	hooks	https://code.claude.com/docs/en/hooks-guide
Permission rule glob syntax (`Bash(npm )`, `Read(src/*)`, `mcp__server__tool`)	—	https://code.claude.com/docs/en/permissions
Programmatic / CLI usage	headless	https://code.claude.com/docs/en/headless
/loop, cron, reminders	scheduled-tasks	https://code.claude.com/docs/en/scheduled-tasks
Push events into sessions	channels	https://code.claude.com/docs/en/channels
Dynamic context injection (`!`command``)	dynamic-context-injection	https://code.claude.com/docs/en/skills#inject-dynamic-context
Complete skill-building guide	skill-complete-guide	https://resources.anthropic.com/hubfs/The-Complete-Guide-to-Building-Skill-for-Claude.pdf
Teleport & Remote Control	session-mobility	https://code.claude.com/docs/en/remote-control
Git worktrees for parallel work	worktrees	https://code.claude.com/docs/en/common-workflows#run-parallel-claude-code-sessions-with-git-worktrees
/batch fan-out for large changesets	batch	https://code.claude.com/docs/en/common-workflows
Chrome extension & Desktop preview	frontend-verification	https://code.claude.com/docs/en/chrome
Loop patterns & stateful automation	loop-patterns	https://code.claude.com/docs/en/scheduled-tasks

name: claude-code-best-practices description: Reference for Claude Code features and best practices: writing CLAUDE.md, skills, hooks, sub-agents, agent teams, headless mode, scheduled tasks, worktrees, batch, frontend verification. Use when authoring/auditing CLAUDE.md, configuring Claude Code, writing skills, or orchestrating agents. Keywords CLAUDE.md, skill, hook, sub-agent, agent team, headless, worktree, /loop. allowed-tools: Read, Grep, Glob, WebFetch, WebSearch, Bash(cat ), Bash(gh ), Bash(git ), Bash(ls )