Market Top Detector assesses tactical market top risk over a two to eight week horizon. Claude collects a few breadth and sentiment readings from the web, runs a Python script that pulls index, ETF and volatility data from Financial Modeling Prep, and produces a six-component composite score with a risk zone from Green to Critical.
The score is a heuristic. The skill states that a score of 70 does not mean a 70 percent chance of a top, and it reports no comparable historical episode when the data does not support a match. It describes market conditions and does not give investment advice.
What You Get
- A JSON and a Markdown report named market_top_DATE_TIME
- A 0-100 composite score with a Green, Yellow, Orange, Red or Critical risk zone
- Six component scores with the strongest warning signal highlighted
- Data freshness warnings for any manual input older than three business days
- A comparison with historical top patterns and what-if sensitivity scenarios
- A delta against the previous report in the same output folder
Common Use Cases
- Checking whether distribution days are piling up while the index holds
- Seeing whether defensive sectors are outperforming growth
- Tracking whether leading growth ETFs are breaking down ahead of the index
- Producing a recurring top-risk report and comparing it with the last one
What Makes It Unique
It joins three named methods, O’Neil, Minervini and Monty, in one weighted score and keeps its claims modest. The skill spells out that the score is a heuristic and not an event probability, that risk-budget percentages are policy heuristics, and that no calibration results are supplied. It focuses on tactical two to eight week signals for 10 to 20 percent corrections, and a separate Bubble Detector skill covers longer horizons.
Under The Hood
The skill runs a Python script from the command line. The script calls the Financial Modeling Prep API for S&P 500, QQQ, VIX, growth ETF and sector ETF quotes and history, using your API key from an environment variable or a flag, and downloads a public breadth CSV from GitHub Pages. Before the run, Claude uses web search to collect the 50-day breadth percentage and the CBOE put/call ratio, and may fetch a Barchart or YCharts page as a fallback. The script writes JSON and Markdown reports into an output folder and reads the previous report there for the delta. Dependencies come from a requirements file and include a market calendar library and requests.