Soft Constraints

Soft constraints give you the benefits of guidance without the rigidity of absolute rules. They communicate preferences while allowing AI to exercise judgment.

Hard vs Soft Constraints

When to Use Soft Constraints

Style preferences:

• Naming conventions with exceptions
• Code organization patterns
• Comment frequency

Situational guidance:

• "Prefer X unless Y"
• "Typically do X, but Z is acceptable"
• "When possible, follow X"

Quality targets:

• Function length guidelines
• Complexity thresholds
• Performance considerations

Writing Effective Soft Constraints

Clear preference: "Prefer composition over inheritance"

With exception: "Prefer functional components, unless state management requires class lifecycle methods"

With reasoning: "Typically keep functions under 30 lines (for readability), but prioritize clarity over length"

Soft Constraints in Cursor Rules

# Soft Constraints
- Prefer named exports over default exports
- Typically use TypeScript strict mode
- When possible, colocate tests with source files
- Generally avoid comments for obvious code

Balancing Hard and Soft

• Use hard constraints for non-negotiables (security, correctness)
• Use soft constraints for preferences (style, organization)
• Too many hard constraints = rigid, unnatural code
• Too many soft constraints = inconsistent output

The right mix gives AI enough guidance to be helpful while enough freedom to be intelligent.