Update dataset card for first-trigger data contract

README.md

@@ -1,18 +1,17 @@
 ---
 tags:
-- kalshi
-- trigger-strategy
-- numpy
-- analyzer-data
+  - kalshi
+  - trigger-strategy
+  - numpy
+  - analyzer-data
 ---
-
 # tsv1 trigger-strategy analyzer data
 
-Public Hugging Face dataset payload for the v1 GitHub Actions trigger-strategy analyzer runner.
+Public preprocessed data for trigger-strategy analysis.
 
-Intended layout:
+Each supported asset uses this layout:
 
-```text
+```
 preprocessed/<asset>/manifest.json
 preprocessed/<asset>/window_event_start.npy
 preprocessed/<asset>/window_outcome.npy
@@ -21,6 +20,11 @@ preprocessed/<asset>/event_price_1e4.npy
 preprocessed/<asset>/event_count_centi.npy
 preprocessed/<asset>/day_ord.npy
 preprocessed/<asset>/day_window_start.npy
+preprocessed/<asset>/first_trigger.npy
 ```
 
-Only files required by the analyzer are intended to be uploaded here. The GitHub runner downloads the selected asset's raw files directly and reconstructs the local `preprocessed/<asset>/` directory.
+`manifest.json` records the asset, included days, window and event counts, the SHA-256 hash of each NumPy file, the first-trigger schema version, and the explicit buy-price axis.
+
+For schema version 1, `first_trigger.npy` has type `uint32` and shape `[139, windows]`. Each value is the offset of the first event in that window whose price is at least the applicable buy price. `UINT32_MAX` means that the window is empty or has no matching event.
+
+Consumers must pin an immutable dataset revision and verify every file against the manifest before use.