correctness-reviewer

To catch logic errors and edge cases before production, review code to mentally execute branches, trace state changes, and identify off-by-one or boundary issues. This generates a targeted correctness report with concrete reproduction steps and suggested fixes.

Detecting race conditions and invalid state transitions in async operations requires tracing asynchronous flows to identify faulty error propagation and swallowed errors. The review process generates actionable findings with suggested fixes addressing the root causes of these state-management bugs.

Reviewing code for null and undefined propagation involves tracing inputs through code branches to detect invalid state transitions and misrepresented errors. The output provides concrete reproduction steps and root-cause fixes to ensure robust behavior across typical scenarios.

Static analysis can find off-by-one errors in loops by mentally executing the code across branches to trace inputs and boundary issues. The review applies this to typical scenarios like loops and async operations, yielding a concise correctness report with recommended fixes.

Code review works for auditing error handling by detecting swallowed or misrepresented errors and faulty error propagation. It applies review logic to typical scenarios to ensure robust behavior, providing actionable findings with concrete reproduction steps and suggested fixes for root causes.