The Prompt Engineering Guide Nobody Needed Until AI Wrote 75% of Our Code

Six months ago, our team adopted AI coding tools. Productivity jumped 30% the first week. Then something unexpected happened: code quality dropped. Bug reports increased. PR review time doubled. The problem wasn't the AI. It was us. We were prompting like users, not like engineers. Here's the prompting framework our team developed after 6 months of trial and error. It works across Cursor, Copilot, Claude Code, and ChatGPT. The Fundamental Mistake Most developers prompt AI like they're talking to a junior developer: "Build a login page" This gets you a login page. With no error handling. No accessibility. No loading states. No rate limiting. With inline styles. Using a deprecated library. AI gives you exactly what you ask for. If your prompt is vague, the output is generic. The SPEC Framework Every prompt should include four elements: S — Situation (Context) What exists already? What's the tech stack? What patterns does the codebase use? P — Problem (What You Need) What specific thing a