Market Breadth Analyzer downloads two public CSV files hosted on GitHub Pages, a detail file with about 2,500 daily rows since 2016 and a summary file, and scores how broadly the market is participating. Six weighted components feed a composite from 0 (critical weakness) to 100 (broad participation).
The report names the health zone, the strongest and weakest components and any data freshness warnings, and it tracks the score over repeated runs. No API key is needed. It describes market participation and gives no investment advice.
What You Get
- A composite breadth score from 0 to 100 and a health zone from Critical to Strong
- Six component scores: breadth level and trend, 8MA versus 200MA crossover, peak and trough cycle, bearish signal, historical percentile and S&P 500 divergence
- A trend label of improving, deteriorating or stable from a saved score history
- Data freshness warnings when the CSV is more than five days old
- JSON and Markdown reports with data provenance and an optional run ID
Common Use Cases
- Asking whether a rally is broad-based or narrowing
- A daily or weekly check of market participation
- Comparing the score with a chart-based breadth review
- Producing a same-run report that downstream skills can check by run ID
What Makes It Unique
It needs no paid data: the inputs are public CSV files, so a run is fully reproducible. Components with missing data are dropped and their weight is spread across the rest, and the report shows both the original and effective weights. Score history is kept across runs so the report can show direction, and the JSON records the data date, retrieval time and source.
Under The Hood
Claude runs a Python script (market_breadth_analyzer.py) from the skill folder through the shell. The script fetches the two CSV files over the web from GitHub Pages with the requests library, computes the six scores locally and writes JSON and Markdown reports to the output folder. It also keeps a history file of up to 20 earlier scores in the output folder, and the instructions say to create a nested output folder first. No credentials are used. A methodology reference is loaded on first use.