writing-skills

Test-driven documentation for Claude agents is created by converting informal guidance into a structured Skill blueprint with YAML frontmatter. This enforces a concise three-section description and tracks metadata like keywords and dependencies.

Structuring YAML frontmatter for skill writing requires a minimal specification of name and description fields. This ensures proper skill discovery and quality by enforcing a concise, structured body explaining when to apply the skill.

Test-driven development applies to process documentation by guiding creators through a RED-GREEN-REFACTOR testing flow. This approach codifies how to write, test, and deploy skills, ensuring consistency and repeatable quality before deployment.

Validating skills before deployment across multiple agents is supported through a toxicity check and metadata tracking. This verifies that the YAML frontmatter and structured guidance meet the required quality standards.

Progressive disclosure in a skill blueprint organizes heavy references behind lightweight SKILL.md sections. This structure allows Claude agents to load detailed guidance on demand without overwhelming the initial discovery process.

A toxicity check for skill metadata is needed when verifying skills before deployment across agents. It tracks metadata alongside YAML frontmatter and dependencies to ensure the structured guidance meets safety and quality standards.