writing-skills

To write skill documentation that agents follow, draft a SKILL.md file with YAML frontmatter containing name and description, then add concise operational guidance and iterative red-green-refactor verification to ensure instructions hold under pressure.

Pressure-testing for AI workflow instructions is a validation process that applies realistic failure scenarios to drafted skills, catching loopholes and weak descriptions that look correct but fail in practice before deployment.

Yes, skill documentation requires YAML frontmatter with name and description fields to enable agent discovery and provide the deterministic operational context needed before deployment.

The best way to validate reusable AI workflows is applying iterative red-green-refactor verification with supporting references or scripts, testing instructions against realistic failure scenarios to catch edge cases before deployment.

Agent instructions fail in practice when documentation contains loopholes or weak descriptions; applying pressure-testing and iterative red-green-refactor verification catches these issues that appear correct but break under realistic conditions.