trade-analyzelisted

# Full stock research note Produces a research note on one ticker where every figure is computed from real data and machine-checked before the note is shown to anyone. ## The rule this workflow exists to enforce **A number that is not in the snapshot JSON may not appear in the report unless it carries an inline source URL on the same line.** Language models are good at producing text that has the shape of analysis. An RSI reading, a Fibonacci level or a support price can be generated fluently without any price series behind it, and the result is indistinguishable from real analysis by eye. That is the failure mode this pipeline is built to make impossible: the numbers come from `tools/`, the agents only interpret them, and `tools/verify.py` checks the finished note. ## Steps ### 1. Build the snapshot ```bash python tools/snapshot.py TICKER --out reports/TICKER.snapshot.json ``` This computes every indicator from adjusted OHLCV, pulls filed annual financials from SEC EDGAR with accession numbers attached, and scores five dimensions. Read the resulting JSON before doing anything else. Check two fields first: - `data_gaps` - everything that could not be resolved. Each entry is a claim you are **not** entitled to make. - `scores.composite.coverage` - below 1.0 means components were dropped and the weights renormalised. Say so in the note. If the command fails, report the failure. Do not proceed with a partial picture and do not fill the hole from memory. ### 2. Ru