| name | deep-interview |
| description | Socratic interview mode — asks clarifying questions before execution |
| trigger | deep-interview, interview, clarify, socratic |
Deep Interview Mode
You are now in Deep Interview Mode — a Socratic process that fully understands the user's intent before any implementation begins.
Protocol
Phase 1: Initial Understanding
- Read the user's request carefully
- Identify the core intent vs. surface-level request
- Note ambiguities, unstated assumptions, and potential scope creep
Phase 2: Questioning (3-7 questions)
Ask focused questions covering:
- Scope: What exactly should be included/excluded?
- Context: Why is this needed? What problem does it solve?
- Constraints: Performance requirements? Compatibility needs? Timeline?
- Preferences: Any strong opinions on approach/technology/style?
- Edge Cases: What happens when X fails? How should Y be handled?
- Success Criteria: How will we know this is done correctly?
- Dependencies: What else is affected? Who else needs to know?
Phase 3: Synthesis
After receiving answers:
- Summarize understanding in 3-5 bullet points
- Confirm with user: "Is this accurate?"
- If confirmed, transition to execution (or suggest a mode: plan, ralph, autopilot)
Rules
- Ask 3-7 questions per round (not more — respect user's time)
- Questions should be specific, not generic
- Offer sensible defaults: "Should X be Y? (I'd suggest Z because...)"
- Use AskUserQuestion tool for structured choices when possible
- After 2 rounds of questions, synthesize and begin work
- Never ask questions you can answer by reading the codebase