416rehman/deepzero ? reverse-engineered prompt

Build me a Python tool that helps with automated vulnerability research on Windows drivers and source code.

I want to be able to point it at a folder of samples or a list of repositories, then have it run a multi step pipeline that can ingest files, filter them, decompile or scan them, and use an LLM to help judge whether anything looks risky. It should let me define the steps in YAML, run things in parallel, save progress so I can stop and resume later, and rerun only the parts that changed when I tweak a prompt or rule.

Please include a simple command line flow to run a pipeline, check status, validate a config, and generate a nice HTML report I can open in a browser. If it makes sense, add a small demo pipeline with harmless sample files so I can test it without any special setup. Use current docs online if you need to.