SkillCT Benchmark Evaluates Agent Skills Across Nine Dimensions

SkillCT moves beyond downstream task execution to assess agent skills through static specification quality, dynamic behavioral robustness, and human-alignment dimensions. The execution-free baseline allows for scalable assessment without requiring task-specific environments or repeated code execution.