testing-skills-with-subagents

Deploy sub-agents to test skills across baseline and skill-enabled runs with pressure scenarios. This approach verifies skills resist rationalization, captures failures systematically, and enforces test-driven refinement before production use.

Pressure scenarios are challenging conditions designed to expose weaknesses in skill performance. Testing applies 3+ distinct pressures, documents how skills rationalize failures, and iteratively refines the skill until it performs reliably under all conditions.

Yes. Sub-agents generate diverse test cases and scenarios tailored to your skill's domain, execute the skill against those cases, and measure accuracy, speed, and resource usage to verify correctness and efficiency.

Sub-agents provide scalable, systematic test coverage across multiple scenarios and pressure conditions. They enforce test-driven discipline, explicitly record rationalizations and failures, and enable rapid iteration without human bottlenecks.

Baseline runs test without the skill to establish a control performance level. Skill-enabled runs execute with the skill active, allowing direct comparison to quantify the skill's actual contribution and identify performance gaps.

Record explicit rationalizations—explanations the skill offers for failures—and analyze them across all pressure scenarios. This documentation drives iterative refinement, surfaces systematic weaknesses, and ensures skills meet compliance standards before deployment.