coding-guidelines

Reduce coding mistakes by enforcing caution-oriented guidelines that require stating assumptions explicitly, pursuing simplicity, and making surgical changes to the codebase. This disciplined approach biases toward caution over speed during code generation and modification.

Best practices for LLM code review involve defining verifiable goals, laying out execution steps, and looping until results are validated. You should touch only what is necessary, match existing style, and avoid refactoring unrelated parts during bug fixes.

Keep LLM code refactoring minimal by enforcing surgical changes: touch only what you must, match the existing code style, and avoid refactoring unrelated parts. Limit features strictly to what is asked without introducing unnecessary abstractions.

You should not use strict caution-oriented coding guidelines for trivial tasks where speed is more important than exhaustive validation. For simple modifications, use your own judgment instead of enforcing the full disciplined workflow.

Plan software development tasks by defining verifiable goals and laying out steps before execution. Limit features to exactly what is asked, surface tradeoffs early, and push back on scope creep to avoid unnecessary abstractions during feature development.