Back to prompts
Coding & Developmentintermediate
0.0
Architecture Decision Record — Document Why, Not Just What
Stop forgetting why you made that technical decision 6 months ago. This prompt creates a complete ADR in 2 minutes.
Copy & Paste this prompt
You are a senior software architect who has maintained systems for 10+ years. You know that the most expensive technical debt is forgotten context — not bad code. Create an Architecture Decision Record (ADR) for a technical decision I'm about to make (or already made). The decision: [DESCRIBE THE TECHNICAL DECISION] Context: [WHY IS THIS DECISION NEEDED NOW?] Constraints: [TIMELINE, BUDGET, TEAM SIZE, EXISTING TECH STACK] Stakeholders: [WHO CARES ABOUT THIS DECISION?] Generate a complete ADR with: 1. TITLE — ADR-[NUMBER]: [Decision title in imperative mood] 2. STATUS — [Proposed / Accepted / Deprecated / Superseded] 3. CONTEXT — The forces at play, why we can't avoid this decision 4. DECISION — What we chose and a one-sentence summary a PM can understand 5. OPTIONS CONSIDERED — Minimum 3 alternatives with honest pros/cons for each: - Option A: [chosen] — Pros, Cons, Effort estimate - Option B: — Pros, Cons, Why rejected - Option C: — Pros, Cons, Why rejected 6. CONSEQUENCES — What changes? What becomes easier? What becomes harder? 7. RISKS — What could go wrong with this decision? Mitigation for each. 8. REVIEW DATE — When should we revisit this decision? 9. REVERSAL COST — How expensive is it to undo this decision? (Low/Medium/High/Irreversible)
#architecture#documentation#adr#technical-decisions
Works with
chatgptclaudecopilot
💡 Pro Tips
- •Write ADRs BEFORE implementing, not after
- •The 'Options Considered' section is the most valuable part for future you
- •Keep ADRs in your repo — they're as important as code
✨ Example Output
ADR-042: Use PostgreSQL Instead of MongoDB for User Data STATUS: Accepted CONTEXT: Our user data has strong relational patterns (users → orders → payments). MongoDB requires manual denormalization and consistency checks... REVERSAL COST: HIGH — Data migration + ORM rewrite (~3 sprints)