Putting It All Together

What you've learned

This module moved fast. You went from understanding what Claude Code can see in your folders, to giving it clear instructions, to reviewing its work, to iterating when the result isn't quite right, to managing conversations as they grow, to setting up a CLAUDE.md so it remembers your preferences between sessions.

That's everything you need to hold a productive conversation with Claude Code. Before we move on, let's pull it into something you can glance at when you're in the middle of actual work and need a quick answer.

The conversation cycle

Every interaction with Claude Code follows the same loop:

1. Give an instruction. Describe what you want in plain language. Focus on the outcome, not the steps.
2. Review the result. Check what Claude Code did. Open the file, spot-check the numbers, compare to what you expected.
3. Iterate if needed. Tell Claude Code what to change. Be specific about what's wrong, not that something is wrong in general.
4. Move on or start fresh. When the result is good enough, move to your next task. If things went sideways, clear the conversation and try again with a better prompt.

This cycle works whether you're cleaning up a spreadsheet, generating a report, or organizing files. It becomes second nature faster than you'd expect.

When to iterate vs. start fresh

Knowing when to keep going in a conversation and when to start over is one of the most useful things you can take from this module.

Keep iterating when:

• The result is close to what you wanted
• You're making small adjustments (formatting, phrasing, a few numbers off)
• You've only corrected Claude Code once or twice
• The conversation context is still relevant to what you're doing

Start fresh with /clear when:

• You've corrected the same issue more than twice
• The conversation has drifted far from your original goal
• You're switching to a completely different task
• Claude Code seems confused or keeps repeating the same mistakes

The two-correction rule from earlier is the simplest version of this. If you've told Claude Code the same thing twice and it still isn't getting it right, clear the conversation and write a better initial instruction. A clean start with a clearer prompt almost always beats a long conversation full of corrections.

When to use each command

Here's a reference for the conversation management commands from this module:

You don't need to memorize this table. Bookmark this page, or better yet, add a reminder to your CLAUDE.md: "When in doubt about conversation management, the summary is in Module 2, page 8."

The five-command starter kit

If you take one thing from this module, make it these five commands:

1. /clear — Start fresh. Use it between unrelated tasks.
2. /compact — Squeeze a long conversation into a shorter summary. Use it when things slow down.
3. Press Esc twice — Rewind. Undo changes to your conversation, your files, or both.
4. /rename — Name your session. Use it when you'll want to come back later.
5. claude --continue — Pick up where you left off.

These five handle most of what you'll run into day to day.

How it all connects

The skills in this module aren't isolated tips. They stack.

Clear instructions (page 2) mean less iteration. Good review habits (page 3) catch problems before they snowball. Specific feedback (page 4) gets you to the right result without going in circles. The worked example on page 5 showed what that looks like in practice — catching a bad assumption, naming the problem, tightening the rules. Conversation management (page 6) keeps each session focused so Claude Code doesn't lose the thread. And CLAUDE.md (page 7) means you stop wasting the first few minutes of every session re-explaining your project.

The payoff is that getting any one of these right makes the others easier. When your instructions are specific and your CLAUDE.md handles the background context, you spend less time correcting and more time on the work you actually care about.

A starting workflow

Here's a workflow worth trying as you put these skills together. It's not a strict procedure — adapt it as you figure out what fits your tasks.

Before your first session on a project:

• Create a project folder with your files
• Type /init to generate a starter CLAUDE.md
• Edit the CLAUDE.md with your project details and preferences

At the start of each session:

• Open the terminal and navigate to your project folder
• Type claude to start (or claude --continue to resume)
• Give your instruction using the outcome-first approach from page 2

During the session:

• Review each result before moving on
• Iterate with specific feedback when needed
• If you've corrected the same thing twice, /clear and rewrite your instruction
• Use /compact if the conversation gets long

At the end of the session:

• Use /rename to name the session if you'll want to come back
• Update your CLAUDE.md if you learned something new about your project (or tell Claude Code to "remember" it)

After a few sessions, you'll find your own rhythm. Some people /clear between every task. Others let conversations run longer and /compact when they need room. There's no wrong approach as long as you're checking results and keeping sessions focused.

What's ahead

With these conversation skills down, you're ready for the practical half of the course.

Module 3 is about working with data: opening CSV files, asking questions about your numbers, cleaning up messy spreadsheets, and generating charts and reports. For many people, this is where Claude Code starts paying for itself.

Everything from this module applies directly. You'll describe what you want from your data in plain language, review the results, iterate when the numbers don't look right, and /clear when you need a fresh start. Same skills, more interesting problems.