Claude Code Skills vs Commands: The Key Difference
You will turn a one-off Claude Code request into a custom slash command, then into a skill, and you will know when each one fits. The example is generating 10,000 fake people with the Faker library.
Was Sie brauchen
Claude Code installed and open in a project folder
A repeated task you want to stop retyping (the video uses fake data generation)
Schritt für Schritt
Run the task once as a normal request
Ask Claude Code to do the job in plain words and let it finish. The video uses Faker, a library that generates mock data for analysis tests. Check that the output is what you want before saving it as anything reusable.
First requestCan you please generate ten thousand fake entries of just random fake people using the Faker library for me, please?
Turn it into a custom slash command
Once the task works, ask Claude to save it as a command. Say explicitly that you want a command, not a skill. Claude reviews what it just did and writes one long prompt (a mega prompt) that reproduces the same output every time.
Create the commandCan you please create a custom slash command for this use case? Make sure it's a command and not a skill, though.
Find the command file
Open the .claude folder in your project. Commands are stored there as markdown files (plain text files), one per use case. Open the new file (in the video, generate fake people) to read the instructions Claude wrote.
Restart Claude Code and run the command
Close and restart your Claude Code session, because a new command does not show up until you do. Type /generate and pick the fake people command from the list. Claude repeats the same job without you explaining it again.
Convert the command into a skill
Ask Claude to turn the command into a skill. Skills are also markdown files, but they live in a separate skills folder and can keep the code they use, so Claude reuses the same script instead of rewriting it each time.
Convert to a skillCan you now turn this custom slash command into a whole skill, please?
Check the skill description
Open the new skill file and read the skill description at the top. Claude always keeps this short description in its memory (the context window), so it knows the skill exists without loading the whole file. Make sure it clearly says what the skill does.
Call the skill with vague phrasing
Test it without a slash. Describe the need in everyday words, even inside a bigger task, and Claude picks the matching skill on its own.
Vague request that triggers the skillHey, I'm going to need help generating some fake people information.
Choose skills or commands for new work
Use a skill when you want Claude to find it automatically, share it with coworkers (including via API), or let agents use it. Use a command only when it must never run by accident, since it only runs when you type its name. My advice: stick to skills going forward.
Worauf Sie achten sollten
New custom slash commands do not appear until you restart your Claude Code session.
A command is only a markdown file, so Claude can get confused and regenerate the script each run. A skill saves and reuses the same code.
Skills can be triggered by vague phrasing, which is useful but can fire when you did not intend it. Commands cannot be invoked by accident.
Agents cannot use custom slash commands, only skills, and Anthropic is moving everything to skills.
