claude /run: Verify It Works by Running the Real App

You've probably had it once: tests are all green, but when you actually launch the app, the screen breaks. /run is a command that directly runs the app to verify whether a code change works properly. The key is verifying with a real run the problems that automated tests alone can't catch.

Definition (What It Is)

/run is a command that directly runs the app to see whether code changes work in the real app. It doesn't stop at running tests (verification code written in advance); it launches the app itself to check actual behavior.

Tests quickly verify "does the code run correctly under preset conditions," but they don't catch everything, like the actual screen appearance or the flow when a user clicks through it themselves. /run fills that gap by actually running the app to verify even problems you can confirm with your eyes. Because it deals with execution flow, it's classified as advanced.

How to Use It (By Difficulty)

Basic — Checking Execution

After changing the code, type this in the input box.

/run

Claude Code launches the app and checks whether it works. You can inspect whether the part you changed actually operates as intended and whether errors occur during execution.

Common Pitfalls

• If Claude Code doesn't know how to build and run your project, /run may get stuck. In that case, if you first make the run method into a skill with /run-skill-generator, /run will then work automatically.
• /run is a real execution, so it takes more time than a test. It's efficient to split usage: tests for quick checks, /run for final verification.
• Running isn't the same as passing. Even if the app comes up, you have to check yourself whether the behavior matches your intent.

Real-World Example

Right after changing the UI (screen) is a representative case. After fixing a button's position or color, launching the app with /run to check whether it looks right on the actual screen lets you immediately spot visual problems that tests don't catch.

Taking It Further

It's also good for checking "does it really work in the flow a user actually uses" after adding a feature. By following the real path from login to payment with /run, you can catch problems at the connecting parts that partial tests easily miss.

Summary

/run is a command that directly runs the app to verify whether code changes work in the real app. You can verify with your eyes even visual and flow problems that tests can't catch. When Claude Code doesn't know the run method, just set it up first with /run-skill-generator.

Reference: Claude Code v2.1.154 (2026.05)