create-agent-skills

To create Claude Code skills and slash commands, you author a SKILL.md file containing YAML frontmatter for metadata and markdown content for instructions, optionally supported by reference files and scripts directories for progressive disclosure.

The best way to structure a SKILL.md file is to keep the markdown concise with essential YAML frontmatter like name and description, then link to detailed reference files in support directories to avoid bloating the main file.

Dynamic context injection in Claude Code skills works by passing arguments and subagent contexts directly into the markdown content, enabling skills to adapt their execution logic based on runtime inputs.

YAML frontmatter in a Claude Code skill should include the name and description parameters, ensuring the description is clear and concise to help the agent understand when and how to invoke the skill.

You need to use progressive disclosure in skill development when your SKILL.md file becomes too large, allowing you to keep the core file concise by linking out to detailed reference files and assets only when needed.

Yes, building custom slash commands requires knowledge of Claude Code syntax and conventions to properly define YAML frontmatter and markdown content formats that the agent can parse and execute.