writing-skills

Skill development uses a RED-GREEN-REFACTOR cycle to build documentation-based skills. You write pressure scenarios exposing agent failures first, then create skill documentation to pass those tests, and finally refactor to cover edge cases.

Agent testing validates process documentation by running subagent-driven pressure scenarios. These scenarios expose skill deficiencies by forcing the agent into complex situations where missing or unclear documentation causes failures.

Pressure scenarios are created by defining subagent tasks that stress specific skill compliance requirements. You design these tests to deliberately trigger agent failures, revealing where the current documentation lacks sufficient guidance for edge cases.

Yes, the TDD approach applies to both new and existing skills. After writing tests that demonstrate an agent's initial refactoring failure, you iterate on the documentation until tests pass, then refactor to handle additional edge cases effectively.

No specific testing frameworks are required as dependencies. Skill compliance is verified through subagent-driven pressure scenarios that evaluate whether the agent successfully follows the documentation to complete the assigned task.

Agents fail when documentation lacks sufficient detail for complex scenarios. Test-driven skill development solves this by first capturing the failure as a test, then refining the documentation until the agent successfully passes the pressure scenario.