32 Claude Code Tricks: From /init to Advanced Power-User Workflows

Key Takeaways

• Run /init on every project — Claude scans the entire codebase and generates a CLAUDE.md file mapping your architecture, conventions, and key files so you never re-explain the project each session.
• /statusline adds a mini dashboard at the bottom of the terminal showing model, context percentage remaining, and running cost — essential for avoiding context rot on long sessions.
• Native /voice command lets you talk directly to Claude Code; still rolling out but also works via any third-party system-wide dictation tool.
• Keep context small: only feed Claude what it needs for the current task. Smaller context windows reduce noise and measurably improve output quality.
• Best hacks are saved for the end of the video — the structure escalates deliberately, so watch through to the advanced section for the highest-leverage techniques.

Commands & Code Mentioned

/init
/statusline
/voice

The Context Management Principle

A recurring theme across multiple hacks in the video is context discipline. Claude performs better with a focused, minimal context window — not because of token limits, but because irrelevant information degrades the quality of responses. The recommended approach is to break large problems into small focused steps, only include files directly relevant to the current task, and use /init-generated CLAUDE.md files as lightweight persistent memory rather than re-pasting the same background on every prompt.

Related on OpenClawDatabase

• Claude Cowork Setup Guide — initial configuration walkthrough including CLAUDE.md setup
• OpenClaw Configuration Guide — deeper dive into project-level settings and memory files
• Commands Reference — full list of slash commands across agent platforms