analyze

A read-only codebase analysis traces concepts across multiple files to explain architecture and behavior without proposing changes. It grounds explanations in evidence and explicitly separates factual findings from inferences.

Evidence-based codebase analysis is a method that grounds explanations of software behavior in traceable references across files. It works by explicitly labeling evidence versus inference, ensuring that architectural explanations are fact-backed and uncertainties are surfaced clearly.

You need cross-file tracing when multiple files or modules influence an outcome, when several explanations exist, or when you must ground your understanding of behavior before making changes. It illuminates architecture and causality across the codebase.

To separate fact from inference during codebase analysis, use a method that requires explicit evidence labeling and traceable references across files. This strict separation ensures that factual observations are distinct from inferred rationale and uncertainties.

Yes, you can use read-only analysis to understand multi-module behavior. It is specifically applied when multiple files or modules influence outcomes, providing a non-destructive synthesis of architecture and causality without altering the codebase.

A limitation of read-only codebase analysis is that it does not propose changes or modifications to the repository. It is strictly non-destructive, meaning it surfaces uncertainties and rationale for behavior but will not generate code fixes or refactoring actions.