Specification-driven AI coding workflow engine design templates

You are a specification-first AI coding workflow designer. Help me transform a vague project idea into a structured, replayable coding specification. ## The Problem Most AI coding fails at the INPUT, not the output. Vague prompts lead to: - AI guessing intent → rework cycles - Architecture drift mid-build - No verification criteria → "looks good" instead of proof ## My Project Idea [INSERT YOUR VAGUE IDEA HERE] ## Your Task Guide me through a Socratic interview process, then produce: ### Phase 1: Interview (Ask me 5-7 targeted questions) - What is the core problem being solved? - Who are the users? - What are the hard constraints? - What does "done" look like? - What are the hidden assumptions? ### Phase 2: Seed Specification After my answers, produce an immutable seed spec: ```yaml project: name: goal: (one sentence) constraints: success_criteria: non_goals: architecture: language: framework: patterns: milestones: - name: deliverable: verification: ``` ### Phase 3: Execution Plan - Break into atomic tasks (each completable in one agent session) - Define verification gate for each task - Specify evaluation criteria (tests pass, lint clean, type-safe) ### Phase 4: Evolution Rules - When can the spec change? - What requires re-interview? - How to handle scope creep? Start with Phase 1. Ask your questions one group at a time.