dry

DRY — Don't Repeat Yourself

Before Applying

If .agents/stack-context.md exists, read it first. Apply this principle using idiomatic patterns for the detected stack. For framework-specific details, use context7 MCP or web search — don't guess.

Principle

Every piece of knowledge in a system should have a single, authoritative representation. When that knowledge changes, you should only need to change it in one place.

Why This Matters in Production

Duplicated logic is a ticking time bomb. When a business rule changes and you update it in one place but miss the copy in another, you get inconsistent behavior that's hard to detect and harder to debug. The more copies exist, the more likely one diverges silently.

DRY is not about eliminating similar-looking code. It's about eliminating duplicated knowledge — the same business rule, the same decision, the same source of truth expressed in multiple places.

Rules

1. Distinguish knowledge duplication from code duplication. Two functions that look identical but represent different business concepts should stay separate. Two functions that encode the same business rule should be unified.
2. Single source of truth for data. Configuration, constants, schema definitions, and validation rules should each live in exactly one place. Everything else should derive from that source.
3. Extract when the pattern is stable. Don't extract on the first occurrence — you don't yet know the right shape of the abstraction. Extract when you've seen the pattern repeat with the same semantics at least twice.
4. Centralize business rules. Tax calculations, permission checks, pricing logic — these must live in one module, not scattered across handlers, frontends, and scripts.
5. Use code generation over manual sync. If two representations must stay in sync (e.g., API types and client types, schema and documentation), generate one from the other rather than maintaining both by hand.

Anti-Patterns

• Shotgun surgery: Changing one business rule requires edits in 5+ files because the rule is duplicated everywhere
• Copy-paste-modify: Cloning a function and tweaking it instead of parameterizing the original
• Parallel hierarchies: Maintaining matching structures in multiple layers (e.g., identical type definitions in backend and frontend that aren't generated from a shared schema)
• Magic strings repeated across files: The same status code, error message, or config key hardcoded in multiple locations
• Documentation that restates the code: Comments or docs that repeat what the code says (and inevitably drift out of sync)