industry-research

Provide an industry or investment theme as the argument, and the workflow builds a causal logic chain, maps upstream to downstream segments, scans global listed companies, and analyzes tier-1 and tier-2 firms with a four-master framework.

It scans US exchanges (NYSE, NASDAQ), A-shares (Shanghai, Shenzhen), Hong Kong stocks, and other international markets such as Japan, Korea, Europe, and Australia, plus industry ETFs and key pre-IPO companies.

Each leading company is scored on business quality (Duan Yongping), moat strength across five types (Buffett), failure scenarios and inversion risks (Munger), and long-term certainty (Li Lu), producing a 1-5 star recommendation rating.

Yes, it outputs a tiered portfolio with core, satellite, and option positions, an ETF alternative, buy/sell/exit signals, and a suggested cap on the theme's share of total portfolio.

After writing the report, a data audit step runs tools/report_audit.py to extract a 15 percent random sample of claims, re-verify each against reliable sources, and issue a pass or reject verdict before publication.

The skill explicitly flags biases such as preference for mature industries, large caps, and English sources. It labels each company with an information sufficiency grade (A/B/C) and marks uncertain conclusions as gray zones rather than fabricating confidence.