aboutcode-org/vulnerablecode-ai-experiments — reverse-engineered prompt

Build me a small Python project that uses an LLM to analyze vulnerability text and turn it into structured data.

I want one simple agent class that can take a vulnerability summary and figure out the package URL, the affected versions, the fixed versions, the severity, and any CWE values it can infer. It should also be able to take a CPE string plus a package ecosystem like pypi and turn that into the right package URL. The goal is to make it easy to feed in messy vulnerability descriptions and get back clean results I can use in code.

Please make it work with any OpenAI compatible API using environment variables or a local .env file, and include a sample env file. Add a couple of runnable examples or basic tests so I can quickly verify each parser method. Keep the interface consistent and easy to use from Python, with clear return values and sensible handling when the text is too vague to extract something. Look up current docs online if you need to.