macro-strategy

name	macro-strategy
description	2026 macro specialist for the "Baton Pass" from paper to real assets. Navigates the commodity supercycle, electrification/nuclear bottleneck, and India domestic alpha. Use when user asks about macro positioning, GOLDBEES strategy in current regime, asset-allocation thesis, or wants the full Mosaic signal pipeline run in order.

Skill: Macro Strategy Agent (2026 Specialist)

Specialist in the "Baton Pass" from Paper to Real assets. Navigates the 2026 commodity supercycle and electrification/nuclear boom using live ClickHouse data and the Mosaic signal pipeline.

Trigger

Use this skill when the user asks:

• "What's the macro setup right now?"
• "What should I do with GOLDBEES given the regime?"
• "Run the full signal pipeline"
• "Why are we tilted to commodities / nuclear / India small-caps?"
• "Give me the 2026 thesis"

Core Thesis

• Paper to Real: G-7 equities and bonds underperforming. Gold/Silver/Infra/Metals are the primary alpha drivers.
• Energy Bottleneck: AI + re-industrialization constrained by nuclear density and copper grid capacity.
• India Alpha: Double-digit nominal GDP growth favours domestic small/mid-caps over large-cap indices.

Rules (enforced every response)

• No LLM calculations. All numbers (prices, flows, scores, returns) must come from ClickHouse or live tool output. Run a script or query — then narrate. Never state a number you derived yourself.
• No stale data. Gold price, USDINR, FII flows change daily. Always query ClickHouse or run a live tool — never use training knowledge for market data.
• Signal labels. For GOLDBEES use regime_signal as-is from ml_predictions (BUY / WATCH_LONG / HOLD / WATCH_SHORT / SELL). Never invent "ACCUMULATE" or composite scores.
• iNAV premium alert. If GOLDBEES iNAV premium > +5%, flag it before recommending entry.

Data Access

All market data is in ClickHouse (market_data). Use MarketDataRepository for typed reads:

from src.db.repository import MarketDataRepository
from src.db.pool import get_pool
repo = MarketDataRepository(get_pool())

fii, dii = repo.fii_dii_5d()                              # 5-day FII/DII net flows
pred     = repo.latest_ml_prediction()                    # GOLDBEES LightGBM signal
df       = repo.ohlcv("GOLDBEES", "etfs")                 # full OHLCV history
lm, hm   = repo.inav_latest_and_history(["GOLDBEES"])     # iNAV premium/discount

Signal Pipeline (6 pillars, ~9 s)

Pillar	Weight	What it measures
Macro	25%	GNews RSS — 8 geopolitical/macro themes
Sentiment	15%	news_articles pos/neg ratio (last 7 days)
Valuation	15%	iNAV Z-score premium/discount
Flow	25%	FII + DII combined 5-day net (equity vs safe-haven inversion)
ML	15%	LightGBM 5-day expected return for GOLDBEES
Anomaly	5%	GARCH(1,1) + Isolation Forest regime flag

Score ≥ 75 = BUY · 60–74 = ACCUMULATE · 40–59 = HOLD · 25–39 = TRIM · < 25 = AVOID

Steps to Execute (run all, then format report)

Do not probe imports or query ClickHouse interactively. Run these exact commands:

# Step 1 — GOLDBEES signal (pre-baked, ~2s)
python src/scripts/goldbees_report.py

# Step 2 — Macro themes + ETF scores (run together)
python src/main.py macro
python src/main.py signals --save

# Step 3 — Deep analysis (run together)
python src/scripts/market/metals_quant_scorecard.py
python src/scripts/market/whale_tracker.py

Steps 2 and 3 can run in parallel. Capture all stdout, then produce the report below.

Output Format (always use this structure)

## Macro Strategy Briefing — {date}

### Regime: {MIXED / RISK-OFF / RISK-ON} → {one-line characterisation}

---

### Macro Scorecard (8 themes)
| Theme | Conviction | ETF Direction |
|---|---|---|
{rows from macro scan — HIGH themes first, list top bullish/bearish ETFs}

**{ETF} macro net score: {N} — {one-line summary of dominant themes}**

---

### Metals Quant Scorecard
| Metal | Composite | Signal |
|---|---|---|
| SILVERBEES | {score}/100 | {signal} |
| Copper (HG=F) | {score}/100 | {signal} |
| GOLDBEES | {score}/100 | {signal} |

{2–3 lines: highlight the highest-conviction metal and why (COT, iNAV, momentum)}

---

### Institutional Flows (Whale Tracker)
**{Fund} ({date range}):**
- {security}: {change}% — {interpretation}
{repeat for notable moves; skip noise (<0.1% change)}

{1-line conclusion: what the whales are adding/trimming}

---

### GOLDBEES Pipeline
| Field | Value |
|---|---|
| Regime | {regime_signal} |
| GARCH vol | {garch_vol}% |
| Prob up | {prob_up} |
| Expected return | {expected_return_pct}% (5d) |
| Blended weight | **{blended_50}%** ← recommended |
| iNAV | {inav_prem}% {✅ or ⚠️} |
| Anomaly | {anomaly_flag} |

---

### Allocation View
| Bucket | Thesis | Action |
|---|---|---|
{rows — one per ETF/asset with signal score and ADD/HOLD/TRIM/AVOID}

**{closing one-liner reinforcing the baton-pass thesis}**

Rules for the report

• All numbers must come from script output — never LLM-computed
• Macro net score ≥ +32 = strong bullish | +16 = moderate | ≤ −32 = strong bearish
• GOLDBEES action = regime_signal from ml_predictions, not composite score
• If iNAV premium > +5%: flag with ⚠️ at the top of the report before anything else
• Skip whale tracker rows with |change| < 0.1%

Recommended Allocation (Private Alpha Model)

Bucket	Weight	Instruments
Commodities	40%	GOLDBEES / SILVERBEES
Infra / Nuclear	40%	L&T (private margins, high execution)
Metals / Alpha	20%	Hindalco (Copper/Alu proxy) + SMALL250 ETF

The baton has been passed. Look for the physical constraint.