cisco-open/ai-deep-sast ? reverse-engineered prompt

Build me a Python security scanning tool that can run in CI and on my laptop, and catch real code vulnerabilities with both fast rules and AI analysis.

I want a quick mode that uses Semgrep to find common issues and secrets in code and config files, then gives a clear report with severity, OWASP category, CWE, CVSS style score, why it matters, and suggested fixes. I also want a deeper mode that indexes the codebase, sends the important findings to an LLM, and does a more thorough analysis across the whole project.

Please make it work from a simple command line, support scanning a file or a folder, let me change the severity threshold, and output readable markdown plus JSON and JUnit XML for CI. Include a config file, Docker support, and a Jenkins pipeline style flow where the fast scan always runs and the AI scan only runs when needed. If you need to check current docs for Semgrep or the model tooling, look them up online.