Prompt Versioning: Guia de Teste e Versionamento de Prompts

Crie um prompt útil e interessante para ferramentas de IA, inspirado em prompts versioning e testes de prompts. Use como referência o conteúdo abaixo para orientar a construção: Título: Prompt versioning - how are teams actually handling this? Descrição: Work at Maxim on prompt tooling. Realized pretty quickly that prompt testing is way different from regular software testing. With code, you write tests once and they either pass or fail. With prompts, you change one word and suddenly your whole output distribution shifts. Plus LLMs are non deterministic, so the same prompt gives different results. We built a testing framework that handles this. Side by side comparison for up to five prompt variations at once. Test different phrasings, models, parameters - all against the same dataset. Version control tracks every change with full history. You can diff between versions to see exactly what changed. Helps when a prompt regresses and you need to figure out what caused it. Bulk testing runs prompts against entire datasets with automated evaluators - accuracy, toxicity, relevance, whatever metrics matter. Also supports human annotation for nuanced judgment. The automated optimization piece generates improved prompt versions based on test results. Formule o prompt de forma que, ao ser executado, ele produza um objeto JSON com os campos: title content description ai_tool difficulty_level rating is_featured tags e topic. A saida deve conter apenas esse JSON, sem texto adicional. A parte content do JSON deve conter o prompt que descrevi acima e deve estar pronto para uso. Indique também qual ferramenta de IA é mais adequada para executar esse prompt entre as opcoes listadas (chatgpt, claude, deepseek, gemini, copilot, perplexity) e inclua a justificativa. Atribua nota de 1 a 5 estrelas, defina o nivel de dificuldade entre beginner, intermediate ou advanced, determine is_featured true ou false, proponha tags relevantes e escolha um topico.