coding-guidelines

To reduce common LLM coding mistakes, enforce behavioral guidelines that require stating assumptions, favoring minimal implementations, and making surgical changes to avoid altering unrelated code. Apply these rules during writing, modifying, or reviewing code to prevent errors.

Best practices for LLM code review involve thinking before coding by surfacing tradeoffs, defining verifiable goals, and executing surgical edits. This disciplined approach ensures refactoring only alters necessary code without introducing unnecessary abstractions.

Coding guidelines are not designed for architecture design, documentation, or non-code tasks. They are strictly applicable to implementation tasks, code changes, bug fixes, and feature development across personal projects or production work.

To enforce simplicity and minimal abstractions in LLM generated code, apply guidelines that favor straightforward implementations. Explicitly state assumptions and ask questions when unclear to ensure the generated code remains minimal and goal-driven.

You need a structured plan for multi-step coding tasks to ensure goal-driven execution. Defining verifiable goals upfront guides the LLM through complex code changes, refactoring, or bug fixes systematically, reducing implementation errors.