fix: chg 01 stage context

was: did too much work, dup'ping some
based on how many tickers.

now: just do the work once

workflows/scan-watchlist-icm/stages/01_broad_filter/CONTEXT.md

@@ -1,4 +1,8 @@
-# Stage 01 Broad Filter: Initial Ticker Screening
+# Stage 01 Broad Filter: Curated List Triage
+
+Note: the input watchlist is always operator-curated (typically 5-50 tickers), never
+a broad-market universe. This stage is a lightweight triage pass on that list, not
+a market-wide scan — don't run Finviz-style universe screens here.
 
 ## Inputs
 - Layer 4 (working): ../00_clarify/output/user_preferences.md
@@ -6,13 +10,16 @@
 
 ## Process
 1. Read user preferences to determine bias, time horizon, and asset class focus.
-2. Read the watchlist filename from user_preferences.md, then load the full ticker list from that CSV in the shared/ folder.
-3. Establish market context by pulling snapshot data for SPY and QQQ to determine the day's regime (trend, range, key levels).
-4. Perform a batch scan (using Finviz or equivalent) filtering for:
-   - Price within ~3% of the 50-day MA
-   - Consolidation/base pattern or pullback setup
-   - Above/below VWAP context
-5. Record the resulting shortlist of candidates (expected 10-15 tickers) for further analysis.
+2. Read the watchlist filename from user_preferences.md, then load the full curated ticker list from that CSV in the shared/ folder.
+3. Establish market context by pulling snapshot data for SPY and QQQ to determine the day's regime (trend, range, key levels). Record this regime read in the output so stage 02 can reuse it instead of re-pulling.
+4. Pull batch snapshot data (price vs 50-day MA, VWAP position, basic volume/range) for every ticker on the list in a single batch request — not one ticker at a time.
+5. Decide whether triage is needed, based on list size:
+   - **List ≤15 tickers:** skip filtering. The operator has already curated it — pass the full list through unchanged as the shortlist.
+   - **List >15 tickers (up to ~50):** apply lightweight triage to cut to a working shortlist (target ~10-15) using:
+     * Price within ~3% of the 50-day MA, or a clear pullback/base pattern
+     * VWAP alignment with the stated bias (above VWAP for longs, below for shorts; either if both)
+     * Drop tickers with no clear structure / no signal
+6. Record the resulting shortlist, with a one-line reason for any ticker dropped in the >15 case.
 
 ## Outputs
-- shortlist.md -> output/
+- shortlist.md -> output/ (includes the SPY/QQQ regime notes from step 3)