xbtlin/ai-berkshire — reverse-engineered prompt

Build me a Claude Code based value investing research framework that feels like having a small AI research team instead of just asking one chatbot for a vague opinion.

I want it centered on four investor lenses, Buffett, Munger, Duan Yongping, and Li Lu, with multi agent parallel research and a final lead analyst style summary. It should be able to do deep single company research, earnings review, management deep dives, industry scanning and funnel screening, quick investment checklists, portfolio review, thesis tracking after buying, fast news reaction, and a Duan style Q and A mode.

The output should be decision ready, not wishy washy. For each company, force a clear pass, no pass, or gray zone conclusion, with confidence, risks, valuation range, and different suggestions for aggressive, steady, and conservative investors.

Please make the reports structured and reusable so I can compare companies over time. Important numbers should be cross checked from multiple sources and calculated carefully, not hand waved. If needed, look up current Claude Code command docs online and wire this up in a way that is easy to run and extend.