finance-investment-team

Invoke the skill with a company name as the argument. It creates a research team, launches four background agents in parallel covering business, financials, industry, and risk, then consolidates their reports into a final investment report with scores and recommendations.

Each agent applies a distinct value-investing lens: business model and moat analysis, financial statements and valuation, industry landscape and competition, and risk and management assessment. The team lead synthesizes all four into a final report with a Buffett-style checklist.

Background agents cannot show interactive permission prompts, so blocked WebSearch silently degrades them to training-knowledge-only answers. The skill checks the permissions whitelist first and stops with instructions if WebSearch is not allowed.

The report_audit.py script extracts financial data points from the Markdown report, randomly samples 15%, and compares each against fetched source values. Points deviating more than 1% fail; any failure returns a reject verdict requiring correction and re-review.

Yes, scripts/twstock_data.py queries the FinMind API for Taiwan stock quotes, valuation, financials, monthly revenue, and dividends. It works anonymously with rate limits or with a token from the FINMIND_TOKEN environment variable.

Avoid it for single-perspective quick analysis or simple information lookups, as stated in the skill description. The four-agent workflow is heavyweight and takes several minutes, so it suits deep research rather than fast fact retrieval.