Skip to content
Docs

Already using Claude

Section titled “Already using Claude”

This page targets Claude Code (CLI) users — not just claude.ai. If you haven’t installed Claude Code yet, start with Setup & configuration.

You don’t need the beginner tutorial. What you want is to quickly spot what you’re missing — and go straight there.

What you probably already know

Section titled “What you probably already know”

Skip these pages if you recognize yourself in each point:

  • Basic prompting — role, context, task, format. You don’t need it explained.
  • Simple CLAUDE.md — you know it exists and use it to keep context between sessions.
  • Claude Code installed — you run claude in your terminal, you know it reads your files.
  • Basic agents — you’ve seen Claude spin up an autonomous subtask or used plan mode.

If you tick all that, jump straight to the next section.

What you might not know yet

Section titled “What you might not know yet”

Points often missed by experienced users:

Sandboxing and fine-grained permissions Claude Code can be configured to only access certain tools or directories. Most people leave default permissions without knowing they can tighten them per project — or harden them for sensitive code. → Permissions & security

exit 2 in hooks exit 1 blocks an action. exit 2 blocks and injects text into the conversation — that’s different. Most people only know exit 0 and exit 1, and miss the ability to enrich context from a script. → Hooks & automation

Passing data between agents In a multi-agent workflow, agents don’t talk directly to each other. Data flows through temp files, env variables, or artifacts. If you haven’t designed a serious multi-agent workflow yet, that’s where it gets complex. → Multi-agents

Managing context in long sessions After 20-30 exchanges, Claude processes a lot of context that’s no longer relevant — responses drift, quality drops. /compact compresses history without losing the thread. Knowing when to use it (before saturation, not after) meaningfully changes long session quality. → Costs & tokens

Section titled “Recommended accelerated path”

If you have time for a serious pass through the docs, here’s the order that makes sense for an experienced profile:

  1. Hooks & automation — guardrails + automation, including exit 2
  2. Agents — worktrees and isolation
  3. Multi-agents — parallel and sequential orchestration
  4. Advanced CLAUDE.md patterns — the layer that makes it all durable
  5. Permissions & security — harden what you’ve built
  6. Costs & tokens — manage consumption in agent-heavy workflows
  7. BMAD — if you want a complete methodology for a real project

You can also jump directly to a page if you know what you’re missing.