agentic-code-reasoning
by KunihiroS
Use this skill whenever code behavior must be reasoned about without executing it — especially when code is provided and needs careful tracing. Invoke it to: trace what a function or program returns step by step, identify which exact operation causes a test failure or wrong value, determine whether two implementations are behaviorally equivalent, simulate an algorithm's execution order manually, or audit for security vulnerabilities like SQL injection or race conditions. Use any time someone asks "what does this code return?", "which step caused the failure?", "are these equivalent?", "what's the execution order?", or "is this vulnerable?" — even when the answer seems obvious without a skill.