A Python agent built on the Swarms framework that reviews code you give it for vulnerabilities an attacker could actually use. Point it at a file from the command line and it reads the code (JavaScript, TypeScript, Python, Go, Java, PHP, Ruby or C#) and reports each problem with the line, the request or input that exploits it, what the attacker gains, the CWE number and a corrected snippet. It ends with a verdict (ship, fix first or do not ship) and a list of what it could not check because the files were not provided. You add your own API key for OpenAI, Anthropic or Swarms. Includes agent.py and the editable system prompt.