llm-eval-promptfoo Prompt-driven LLM testing using Promptfoo. Evaluate GPT, Claude, or local models across accuracy, safety, and bias — beginner-friendly, CI-ready.