C.L.E.A.R. Review Framework

The C.L.E.A.R. framework provides a structured approach to code review when working with AI. It addresses the most common failure modes in AI-generated code.

The Framework

C — Context

Does the code understand its full context?

• Integration with existing code
• Project conventions and patterns
• Dependencies and imports
• Where this code fits in the larger system

L — Logic

Is the logic correct and complete?

• Control flow makes sense
• Conditions are properly structured
• Return values are correct
• No logical contradictions

E — Edge Cases

Are edge cases handled?

• Empty inputs
• Null/undefined values
• Boundary conditions
• Error states
• Concurrent access

A — Assumptions

What assumptions did AI make?

• About input format or type
• About environment or configuration
• About error handling upstream
• About performance requirements

R — Requirements

Does the code meet actual requirements?

• Functional requirements satisfied
• Non-functional requirements (performance, security)
• Matches the original intent
• Nothing extra or missing

Using C.L.E.A.R.

Run through each checkpoint mentally or explicitly:

1. Quick scan for obvious issues
2. C.L.E.A.R. review for anything non-trivial
3. Focus extra attention on areas AI commonly misses (E and A)
4. Document assumptions for future reference

The framework takes 30-60 seconds but catches issues that would take hours to debug later.