SkillEngenhariaCriação

Generate Synthetic Data

Generate diverse synthetic test inputs for LLM pipeline evaluation using dimension-based tuple generation.

Em:LLM System Evaluation Loop
Ações
PerfilDev
ProfundidadeMédia
Idiomaen-US
Objetivo

Em uma frase.

Use this skill when bootstrapping an eval dataset, when real user data is sparse, or when stress-testing specific failure hypotheses. The skill builds tuples across application-specific dimensions, converts them to natural-language queries in a separate prompt, and filters for quality - producing ~100 diverse traces ready for error analysis.

Constelação

Onde
ela vive.

Workflows que usam
Bundles que incluem