The biggest lever isn’t the model — it’s the context
Claude Code ships with dozens of slash commands. Knowing them lets you steer cost, depth and pace deliberately instead of fighting the tool. Most day-to-day failures aren’t model failures — they’re context failures.
Four questions solve most cases:
- New, independent task? →
/clear— a fresh start beats drift. - Long, but you need the history? →
/compact— condenses instead of dropping. - Took a wrong turn? →
/rewind— back to the last checkpoint. - Testing an alternative? →
/branch— fork without losing the main thread.
The session loop
From empty project to review, every session follows the same shape: /init (create CLAUDE.md) → /plan (think first) → prompt (build) → /context (check the fill level) → /compact (condense) → /security-review (scan before PR/merge).
The interactive full version
The complete, interactive cheat sheet — every command by purpose, with aliases and version notes — is free in our knowledge base: Claude Code · Command Cheat Sheet.
Prefer a printout next to the keyboard? The print-optimised one-page version is below as a PDF. If you’re getting your dev setup ready for agentic AI and have questions — we’re happy to book a free demo call.
Read the full article — sign up free
The full article is member content. Magic-link login, no credit card, no risk — and the rest of the article is readable immediately.