writing-skills

To create and test Claude Code skills, use a six-phase workflow from intent capture to deployment, applying frontmatter validation and evaluation tooling to ensure reliable, discoverable outputs.

Standardize skill development by enforcing frontmatter validation, token budgets, and bundled resources, relying on evaluation pipelines to benchmark and prove consistent skill quality across all types.

Yes, the skill writing workflow supports discipline, workflow, reference, and hybrid skill types, providing clear alignment with skill-tdd testing methodology to validate each format reliably.

Evaluate and refine a drafted SKILL.md by running benchmarking and evaluation pipelines to test outputs, then iterating through the optimize phase to ensure token budgets and quality standards are met.

Skills fail frontmatter validation or exceed token limits when structural metadata is incomplete or content is too dense; the workflow enforces validation and token budgets to prevent these deployment blockers.

Yes, bundled references and scripts assist authors and evaluators by standardizing the end-to-end process, ensuring skills remain maintainable, reliable, and easy to discover across the full lifecycle.