Claude Code Workflow Tips After a Year of Daily Use

Practical recommendations from a senior product designer

I’ve been using Claude Code daily for a year now, and I’m a seasoned product designer.

Here’s the 7-step workflow that’s made me genuinely productive after a lot of trial and error:

1. Start project with /init command to create project context for Claude

The first thing I do when I start working on a new project is prompt Claude Code

/init

Claude will create a CLAUDE.md file with contextual information about this specific project.

Example of output that Claude provides in response to /init command for a simple HTML project (landing page).

2. Use “Plan” mode for anything complex

Don’t jump straight to implementation. Instead, create a proper context for AI to get the best possible output. In Plan mode, before Claude writes a single line of code, it drafts its plan which explains its approach and asks you important questions. This saves me a lot of back-and-forth.

Plan mode in Claude Code.

Pro tip for planning with Claude Code: use Codex for planning

Two models are better than one. And the great thing about Claude Code is that you can ‘plug’ Codex in your Claude environment. To do that, you will need to install the Codex plugin

/plugin install codex@openai-codex

and you can use Codex to review Claude’s plan before implementation using the following command:

/codex:adversarial-review

An adversarial reviewer will catch many issues in the plan and save you a lot of tokens you otherwise would spend on unnecessary iterations.

[embed]Codex plugin for Claude Code Why, when, and how you should use it in the product design processuxplanet.org

3. Start small

When we work on a new product design, we rarely build the entire product in one go. Most of the time, we break a large task (creating a new product) into subtasks and handle them one by one. The same approach applies to AI work.

If you ask Claude to “implement the whole product”, Claude will do it, but there is a high chance that many areas of these new features will require refinement and extra iterations. By handling small tasks, you have a lower chance of facing problems.

Pro tip for optimizing Claude Code work: use a visual preview of your solution

The nice thing about Claude Code is that it can show you a preview of your future product. It’s even possible to ask Claude Code to show a specific part of your product if you’re interested in how it works. For example, in my case, I typically ask Claude Code to visualize a new feature in a product in the context of a particular user flow. The great thing is that the prototype shouldn’t be created in the same tech stack as your app. You can ask Claude Code to create a simple HTML prototype to test the feature.

4. Don’t explain bugs in your language

When Claude generates a design for you, and you’re facing an error state during the interaction with your product, don’t try to explain the problem in your own words. Instead, copy-paste the error code and ask Claude to fix the issue. Claude is trained on a massive database of error codes, and it’s often much easier for AI to figure out the root cause of the problem when you provide the raw error code to it.

Pro tip: Ask Claude to perform a security audit of your codebase

Just because your product works as expected doesn’t mean that it’s safe to use. It’s always a good idea to run a security audit before shipping a product.

5. Transfer context to the session when you face context window bloat

If you need to start a new session due to context bloat and don’t want to lose important contextual details, you can ask Claude Code to create a dump with this information.

Here is a prompt I use:

I want to start a new session, and prepare all the required information
for that before I close the current session

Claude will save the session details in MEMORY.md

Claude Code created a dump of the project before starting a new session.

6. Run /simplify before reviewing source code

Claude tends to over-engineer and write more code and more complex code than necessary. This command

/simplify

will make Claude Code to review the source code and optimize it first before the manual review.

Asking Claude Code to simplify the index.html file of the landing page it generated for me.

Summary of /simplify task that Claude Code completed for me.

7. Ask Claude to summarize conversation at the end of each session

Similar to how product teams that practice weekly sprints do a retrospective of their work after a sprint, it makes sense to do a retrospective with Claude at the end of each session.

You can send Claude a prompt

What did you learn during this session?

or build a dedicated Claude skill for that (like a custom /retro skill) that will generate the insights Claude learned from the session.

Next, you can analyze and correct the insights Claude gathered. This is a great way to build up institutional knowledge that Claude stores in long-term memory.

Pro tips for Claude retro: create a separate markdown file for Claude insights

First one is for Claude Code:

You can create a separate markdown file for Claude’s retrospective notes in your project (i.e., retro-notes.md) to make it easy both for you and Claude to review and refine the notes.

Such approach also gives you a very handy ability: Claude can use this file to load specific context on-demand.

Second one for you:

For self-improvement, type /insightsin the chat, and Claude will create an HTML file of what Claude thinks you can improve on. It’s actually very good. Do this regularly.

Insights that Claude Code generated for me.

Want to master Claude Code?

Check out my complete guide to Claude Code, packed with highly practical insights on how you can integrate it into your design process. This guide is updated on a weekly basis, so you’ll always get the latest information about Claude Code.

[embed]Claude Code: Practical Guide for Product Designers Practical guide for product designers who want to master Claude CodeIt covers the following topics: Claude Code Best…babich.gumroad.com