🧭 TL;DR: if your model got smarter but your instructions stayed the same, they might now be holding it back. OpenAI shares how to rethink skills and prompts when upgrading to GPT-6 Astra. Title: Rethinking skills and prompts for GPT-6 Astra URL: 📌 Highlights 📝 Write skill descriptions with narrow trigger conditions, not broad ones 📂 Use progressive disclosure instead of front-loading full context 🎯 Overly specific guidance can now hurt results more than help 📄 Tie AGENTS.md references to specific needs, not blanket must-reads ⚖️ Old restrictions written for prior models can over-constrain Astra 🛑 Without explicit completion criteria, it may stop work too early The counterintuitive but practical lesson: as the model gets smarter, your instructions should get simpler, not more elaborate. #AIAgents# #PromptEngineering#