skill-creator

A Claude Skill requires strict YAML frontmatter in SKILL.md with required fields like name and description, followed by markdown body sections. The name must use hyphen-case, descriptions must avoid angle brackets, and you organize supporting scripts, references, and assets in bundled directories to keep context efficient and the skill maintainable.

Claude Skills use three-level progressive disclosure to minimize context bloat: load essential metadata first, then scripts and references on demand, and assets only when needed. This structured approach ensures your skill performs efficiently without overwhelming Claude's context window with unnecessary information upfront.

Yes. Claude Skills support domain knowledge, workflows, and tool integrations through a step-by-step process: define skill requirements, plan the structure, initialize SKILL.md with proper frontmatter, edit sections for your workflow logic, and package bundled resources. This design ensures your automation is robust and AI-ready.

Skill names must use hyphen-case (e.g., data-analyzer, report-generator) and descriptions have enforced length limits to maintain conciseness. These constraints ensure metadata remains scannable and skills integrate cleanly into larger AI systems without naming conflicts or context bloat.

Bundle scripts and references as separate components within your skill directory structure. Scripts handle execution logic while references provide supporting data or documentation. This separation keeps your SKILL.md lean and allows Claude to load only the components needed for each task, improving performance.