Skip to content
Leo.
← All projects

Case study

Stock Analyst

Research briefs that always argue both sides — no scores, no buy buttons.

Runs locally inside Claude Code; briefs are personal research, not investment advice.

A Claude Code skill that turns one command, /brief TICKER, into a two-sided research brief built from free market data: snapshot, bull case, bear case, what would kill the trade, catalysts and risk guardrails.

The problem

The original wish was a tool that finds the best stock with no effort. No tool has a reliable edge on a 2–6 week horizon, so a ‘pick the winner’ button would only give false confidence.

The approach

  • Reframed the goal from prediction to research automation.
  • Every brief must give the bear case the same weight as the bull case, plus an explicit ‘what would kill this trade’ section.
  • Guardrails built in: an earnings-in-window warning and liquidity/volatility checks.
  • Deliberately no scores, ratings or buy/sell signals — and it never executes trades.

The outcome

  • A repeatable research routine that takes seconds instead of an evening.

What's next

  • Backtester before any screener rules
  • Screener
  • Trade journal

Next case study

Investment Tracker

→