Socratic Tutor Prompt Builder

Construct a reusable system-prompt scaffold for a Socratic AI tutor agent — a tutor that asks rather than answers, scaffolds reasoning instead of supplying it, and adapts question depth to the student's response. Output is a deployable system prompt for the school's approved AI tool (Khanmigo Custom, MagicSchool Student-side, SchoolAI Spaces, district-licensed Copilot/Gemini for Education, ChatGPT Edu, or a custom GPT/Project), plus the teacher-facing companion: a scope statement, the off-topic and unsafe-topic stop conditions, the give-the-answer escape hatch, and a sample 4-turn dialogue showing what good and bad responses look like. Designed around the four design principles the Socratic-AI-tutor research literature has converged on: structured questioning patterns, adaptive scaffolding, reflective pauses, and human teacher oversight of transcripts.