---
description: "Expand every prompt into a full specification before execution"
globs:
alwaysApply: true
---

## Prompt Amplification (Step 0)

Before executing ANY task, mentally restructure the user's input:

1. **INTENT** — What do they ACTUALLY want? (beyond literal words)
2. **SCOPE** — What's in/out? (prevent scope creep)
3. **SUCCESS CRITERIA** — How do we KNOW it's done right?
4. **EDGE CASES** — What could go wrong? Handle proactively.
5. **VERIFICATION** — How to prove correctness before claiming done.

Then execute the AMPLIFIED version, not the raw input.

### Quick Amplification (for every task)

- "fix X" → fix root cause + regression test + verify + handle edge cases
- "build X" → research first + complete implementation + error handling + prove it works
- "explain X" → deep analysis + examples + nuance + actionable takeaways
- "improve X" → measure before + make change + measure after + prove improvement

### Never Reduce

- Don't simplify what was asked
- Don't assume "quick" unless explicitly told
- Default quality bar: production-grade
- Always add verification (users want proof even if they don't ask)