Notes about using Claude Code

Today I experimented with Claude to help write code and structure a new Astro blog post. The goal was to keep the content simple, practical, and ready for publishing.

With Claude Code I can use ’#’ to write an instruction and Claude Code saves it to CLAUDE.md. For example:

# Use comments sparingly. Only comment complex code.

Referencing files

To mention a file we can use ’@’. For example, if I want to refer myFolder/myFile.pas I’ll write:

Analyze @myFolder/myFile.pas

Planning mode

To enable Planning Mode press Shift + Tab twice (once if you want auto-accepting edits).

In Planning Mode Claude will:

• Read more files in your project
• Create a detailed implementation plan
• Show you exactly what it intends to do
• Wait for your approval before proceeding

Thinking modes

The available thinking modes include:

• “Think” - Basic reasoning
• “Think more” - Extended reasoning
• “Think a lot” - Comprehensive reasoning
• “Think longer” - Extended time reasoning
• “Ultrathink” - Maximum reasoning capability

Planning vs Thinking

Planning Mode works best for broad tasks involving multiple files, while Thinking Mode suits complex logic and debugging. Both can be combined, but keep in mind they use more tokens.

Commands

/clear

Removes completely the conversation history.

/compact

Summarizes your entire conversation history preserving the learning.

/init

Creates the CLAUDE.md file in your project.

/help

Shows all available commands.

/cost

Displays the tokens consumed in the current session.

/model

Switches the model (Sonnet, Opus, etc.).

/exit

Closes Claude Code.

Custom commands

Go to .claude folder in the project directory and create a new directory commands (if it does not exist). There create a new markdown file with the name of the command we want to create. Write in the markdown file the instructions for the command. If we want to accept arguments we use $ARGUMENTS inside the file.