write-a-skill

To create agent skills that are discoverable, use structured SKILL.md templates with trigger-based YAML frontmatter and progressive content disclosure, ensuring AI systems can correctly load your custom workflows. This approach enforces compliance with usability standards.

Progressive disclosure in skill structure is a content organization method that layers information so AI agents load only what is needed. You need it when building complex agent skills to ensure maintainability and prevent overwhelming the system with unnecessary context.

To bundle utility scripts and reference documents when building custom skills, follow resource bundling guidance to add assets only when needed for deterministic tasks or edge cases, keeping them organized within the skill directory structure alongside the SKILL.md file.

Yes, you can use the guided requirement gathering workflow to build a skill for domain-specific CSV validation and formatting. It helps define the scope, draft the compliant SKILL.md, add necessary validation scripts, and structure reference docs for edge cases.

The best way to define the scope and requirements for a new AI agent skill is through structured requirement gathering prompts that collect details about the domain, specific use cases, and resource needs upfront before drafting the SKILL.md file.

Custom agent skills fail to load or remain undiscovered when they lack proper skill directory structure, trigger-based YAML frontmatter, or clear descriptions, which prevents the AI system from correctly parsing and matching the skill to user intents.