Securing a .NET App with Claude Code

A measured answer to a loose question: how far can you trust Claude Code with the security of an app? A small ASP.NET Core patient portal was written twice, once secure and once with 20 planted vulnerabilities and six decoys, with a hidden suite of attack tests as the judge. Claude Code reviews it, fixes it, rebuilds features under three levels of guardrails and is fed planted instructions; scanners run on the same files; every result is counted, including the ones that went against the guesses written down first.

5 parts · ~36 min of reading, in order, nothing assumed.

← Build with AI All series