Claude Skill Authoring Expert FAQs

Question 1

What is the difference between a command and a skill in Claude Code?

Accepted Answer

Commands and skills are now functionally the same, but a 'skill' uses a directory structure (skills/name/SKILL.md) to support reference files and auto-loading context, whereas a 'command' is typically a single markdown file (commands/name.md) for simple tasks.

Question 2

How can I prevent a skill from using up my context budget?

Accepted Answer

Set 'disable-model-invocation: true' in the YAML frontmatter. This removes the skill description from the 2% context budget, making it a pure slash command that only loads when you explicitly type it.

Question 3

Can Claude Code skills run shell commands?

Accepted Answer

Yes, using Dynamic Context Injection. You can use the '!' followed by a backtick-wrapped command to execute shell tasks and inject the output directly into the skill context before it's sent to Claude.

Question 4

What is the 'context: fork' setting used for?

Accepted Answer

Setting 'context: fork' runs the skill in an isolated subagent. This is ideal for complex tasks like deep research or code reviews where you want the agent to operate independently of the main conversation history.

Claude Skill Authoring Expert

Key Features

Use Cases

Claude Skill Authoring Expert

Key Features

Use Cases