OpenAI's documentation states that GPT-6 Astra asks clarifying questions more often than GPT-5.6 Sol, which can stop work prematurely. The company recommends prompts that instruct the model to infer user intent and show a "bias towards action." Other guidance includes auditing skill files like AGENTS.md, avoiding "slop words" such as "delve," limiting delegation to sub-agents, and rerunning tests only when failures justify it. Recommended prompts are available on the model documentation page.
No score is assigned. Sources and their independence are shown in the citation chain below.