skill-architecture

Claude Code skills use three-tier progressive disclosure: YAML frontmatter for metadata, a SKILL.md file for detailed documentation, and bundled references/scripts/assets for resources. This layered structure ensures portability, security, and clarity across skill distribution and usage.

Validate YAML frontmatter syntax, trigger description integrity, tool allowlisting, and bash compatibility for CLI environments. The architecture provides validation scripts that check these requirements and ensure skills meet functional standards before deployment to ~/.claude/skills.

Bundle scripts, references, and assets as discrete components within the skill structure. This organization pattern supports progressive disclosure, maintains CLI-specific bash compatibility, and enables ongoing maintenance through evolution logs and security practice checks.

Yes. The architecture provides TodoWrite templates for skill creation, updates, and evolution tracking. Templates enforce best practices for YAML frontmatter, progressive disclosure patterns, and resource organization while maintaining security and portability standards.

Apply tool allowlisting, validate CLI bash compatibility, structure progressive disclosure to limit exposure, and maintain security practice checks through validation scripts. These patterns ensure skills operate safely within the Claude Code environment and meet functional requirements.

Use standard YAML frontmatter, bundle all dependencies as components, apply validation scripts for integrity checks, and document evolution through maintenance logs. This approach ensures skills transfer reliably to ~/.claude/skills and function consistently across CLI deployments.