writing-skills

To write and verify production-ready skills using a test-driven process, author a minimal SKILL.md with required YAML frontmatter, run failing subagent pressure scenarios to capture baseline behavior, and iterate with RED-GREEN-REFACTOR cycles to confirm agent compliance.

A pressure scenario in skill development is a failing subagent test run to document agent rationalizations before authoring instructions. You need it when creating operational skills to prevent ambiguous guidance and ensure agent behaviors match your intended deployment outcomes.

To create a SKILL.md file with the right discovery triggers, include YAML frontmatter with the skill name and a discovery-focused description, ensure broad keyword coverage, and apply token-efficiency guidance so agents can reliably find and execute the operational skill.

Agents skip required review steps due to ambiguous guidance or loopholes in instructions. You can prevent it by documenting the failing behavior in a pressure scenario, authoring a minimal counter-instruction in your SKILL.md, and revalidating with subagent verification runs.

The limitations of test-driven skill authoring include the requirement to compose and maintain baseline failing tests for every behavior, and the need for agent-enabled development workflows to run RED-GREEN-REFACTOR iteration cycles for closing loopholes and verifying compliance.

Yes, writing-skills operates without external dependencies, requiring only an agent-enabled environment to run subagent pressure scenarios and iterate on SKILL.md files using the RED-GREEN-REFACTOR cycle to achieve verified agent compliance.