questions-are-not-instructions

To stop an AI assistant from taking impulsive action, use an engagement-first rule that requires it to acknowledge user feedback and ask clarifying questions before executing any instructions. This prevents misaligned responses.

An engagement-first workflow for prompts forces the assistant to answer literal questions and request missing details, stopping before any execution until the user provides explicit direction. This ensures clear communication.

To ensure literal interpretation, apply a stop-before-action rule that requires the assistant to answer questions directly and confirm user intent. The assistant must pause execution until receiving explicit direction.

Yes, you can enforce a hard stop on instruction execution by implementing a rule that prevents the assistant from acting until the user provides explicit direction. This applies to prompts containing questions, feedback, or observations.

Use a clarify-first communication workflow whenever prompts contain questions, feedback, challenges, or observations. This approach prevents misaligned actions by requiring explicit clarification and user direction before execution.

The limitation of an engagement-first rule is that it enforces a strict stop on execution whenever a prompt contains questions or feedback, requiring additional user input before proceeding. This adds extra steps to the workflow.