strip prompt-echo leakage in indicator summaries

DeepSeek occasionally regurgitates the system prompt verbatim
("Constraints: ≤60 words...", "Example good: ..."). Three-pronged fix:

1. Removed the inline good/bad example blocks from the per-group and
   aggregate system prompts — DeepSeek was treating them as templates
   to copy. The hard constraints alone are clear enough.

2. Expanded the LEAK_PATTERNS list to catch the prompt-label echoes
   that still occasionally slip through ("Key observations:", "The
   indicators are:", "Must cite ...", "Should give ...", bare "Key:").
   Cleanup now runs up to 6 passes for compound leakage.

3. Added looks_like_leakage() — if the cleaned output still contains
   tell-tale phrases ("≤60 words", "instructions:", etc.), the summary
   is skipped rather than persisted. Logs a 'leakage_detected' warning
   and an ai_calls row with status=leaked so we can see the failure
   rate over time. The previous good summary stays visible.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

app/services/openrouter.py

@@ -199,20 +199,12 @@ interpretation. Don't list moves; explain them.
 - Multi-week / multi-month horizon. 1-day moves under 2% are noise — skip.
 - No buy/sell language. No predictions. No watch list. No TL;DR. No date \
 header. No "system temperature" line — that belongs to the full daily log.
+- Output the read directly. Do NOT include phrases like "Example", "Good \
+example", "Bad example", "Reference", or any meta-framing of your output.
 
 {tone_block}
 
 {analysis_block}
-
-# Bad example — describes what happened
-"S&P +5.2% 1m and Nasdaq +8.8% 1m diverge from FTSE -3.4% and Euro Stoxx \
--2.6%. The US-vs-rest gap is widening."
-
-# Good example — interprets what it means
-"The US-vs-rest equity gap is funded by AI-capex concentration in 7 names; \
-the breadth-weighted RSP barely keeps pace with SPY, which is the classic \
-late-cycle marker — narrow leadership, not broad recovery. The 5% 1m gap \
-between Nasdaq and FTSE is a narrative trade, not a fundamental one."
 """
 
 
@@ -255,21 +247,12 @@ risk premium is in commodities but not vol". Cite no more than 3 specific \
 numbers, and only as anchors for the interpretation.
 - Multi-week / multi-month horizon. 1-day moves under 2% are noise.
 - No buy/sell language. No predictions of specific levels.
+- Output the read directly. Do NOT include phrases like "Example", "Good \
+example", "Bad example", "Reference", or any meta-framing of your output.
 
 {tone_block}
 
 {analysis_block}
-
-# Bad example — describes
-"Equities are up, real yields are higher, HY OAS is tight, breadth is \
-narrowing."
-
-# Good example — interprets
-"The tape is paying a rising real discount rate (US 10y real +15bp 1m) with \
-conviction for AI growth, but credit refuses to confirm and breadth is \
-narrowing — that combination is what late-cycle looks like, not pre-crash. \
-The risk is not the level but the convergence: if any one of credit, \
-breadth, or vol turns, the others will follow fast."
 """