Create diverse synthetic test inputs for LLM pipeline evaluation using dimension-based tuple generation. Use when bootstrapping an eval dataset, when real user data is sparse, or when stress-testing specific failure hypotheses. Do NOT use when you already have 100+ representative real traces (use stratified sampling instead), or when the task is collecting production logs.
安装
尚未确认此技能的独立目录。下面命令会打开仓库技能选择器，请按上游说明选择技能。
skills CLI
npx skills add ai-evals-course/evals-skills