diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..bd5590b
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,21 @@
+node_modules/
+.npmrc
+.env
+.env.*
+__tests__/
+coverage/
+.nyc_output/
+dist/
+build/
+.cache/
+*.log
+.DS_Store
+tmp/
+.tmp/
+__pycache__/
+*.pyc
+.venv/
+venv/
+*.egg-info/
+.pytest_cache/
+READY_TO_PUBLISH
diff --git a/AGENTS.md b/AGENTS.md
new file mode 100644
index 0000000..cb3c6c4
--- /dev/null
+++ b/AGENTS.md
@@ -0,0 +1,26 @@
+# AGENTS
+
+This repository hosts a production-oriented skeleton for ML-CV Hedge Studio MVP.
+
+- Core DSL: Asset, MarketSignal, RiskState, HedgePlanDelta, GraphOfContracts
+- Engine: HedgeSynthesisEngine (budget-aware hedging planner)
+- Adapters: PriceFeedAdapter, OptionsVenueAdapter (toy builders)
+- DeltaSync: deterministic replay log (delta_sync.py)
+- Tests: tests/test_core.py validates core data structures and flow
+
+How to contribute
+- Install: python3 -m pip install -r requirements.txt (if present)
+- Run tests: pytest
+- Build: python3 -m build
+
+Commands to run in this repo (for agents):
+- Run unit tests: pytest
+- Build package: python3 -m build
+- Replay deltas: python -m idea91_ml_cv_hedge.delta_sync
+
+Architecture overview
+- DSL primitives map to a canonical Graph-of-Contracts approach (Asset, MarketSignal, RiskState, HedgePlanDelta)
+- A lightweight hedge synthesis engine computes cross-venue hedge deltas under budgets
+- Adapters connect to real venues/simulators; Sandbox adapters are provided as MVP
+- DeltaSync ensures deterministic replay across partitions and reconnects
+- Optional privacy-friendly signal sharing via aggregation (stubbed in MVP)
diff --git a/README.md b/README.md
index 35315da..583c49c 100644
--- a/README.md
+++ b/README.md
@@ -1,3 +1,15 @@
 # idea91-ml-cv-hedge
 
-Source logic for Idea #91
\ No newline at end of file
+ML-driven cross-venue hedge synthesis MVP skeleton
+
+This repository provides a production-oriented skeleton for an ML-augmented cross-venue hedge synthesis engine. It includes a canonical DSL, deterministic delta replay, toy adapters, and a minimal hedge engine to bootstrap MVP development.
+
+How to run
+- Install dependencies and tests: see test.sh
+- Run tests: ./test.sh
+
+Architecture overview
+- Core primitives: Asset, MarketSignal, RiskState, HedgePlanDelta, GraphOfContracts
+- Engines and adapters to enable cross-venue hedging experiments
+- Delta-sync for deterministic replay
+- MVP plan in repository: two assets, two venues, two toy adapters
diff --git a/pyproject.toml b/pyproject.toml
new file mode 100644
index 0000000..199b96a
--- /dev/null
+++ b/pyproject.toml
@@ -0,0 +1,30 @@
+[build-system]
+requires = ["setuptools>=42", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "idea91_ml_cv_hedge"  # final packaging must exactly match this or idea91_ml_cv_hedge
+version = "0.1.0"
+description = "ML-driven cross-venue hedge synthesis MVP skeleton"
+requires-python = ">=3.8"
+
+[tool.setuptools]
+# Specify package-data as a mapping from package name to a list of glob patterns.
+# Empty list means no data files to include. This fixes a PyPI build
+# configuration error observed when using setuptools >= 42+.
+[tool.setuptools.package-data]
+idea91_ml_cv_hedge = []
+
+[tool.setuptools.packages.find]
+where = ["src"]
+
+# setuptools-scm integration removed to maintain compatibility with CI
+
+[tool.pytest.ini_options]
+minversion = "7.0"
+addopts = "-ra -q"
+testpaths = ["tests"]
+
+[project.urls]
+Homepage = "https://example.org/idea91_ml_cv_hedge"
+Repository = "https://example.org/idea91_ml_cv_hedge.git"
diff --git a/src/idea91_ml_cv_hedge/__init__.py b/src/idea91_ml_cv_hedge/__init__.py
new file mode 100644
index 0000000..a3de5d7
--- /dev/null
+++ b/src/idea91_ml_cv_hedge/__init__.py
@@ -0,0 +1,19 @@
+"""idea91_ml_cv_hedge: ML-Driven Cross-Venue Hedge Studio (MVP Skeleton)
+
+This package provides a minimal, production-oriented skeleton for the ML-CV Hedge Studio
+project described in the request. It includes:
+- A DSL (Asset, MarketSignal, RiskState, HedgePlanDelta) and a Graph-of-Contracts registry scaffold
+- Two toy adapters (price feed and options venue) to bootstrap interoperability
+- A deterministic delta-sync mechanism for replay across partitions
+- A simple hedge synthesis engine with a tiny optimization placeholder
+- Basic tests to validate core data structures and end-to-end delta flow
+"""
+
+from .dsl import Asset, MarketSignal, RiskState, HedgePlanDelta, GraphOfContracts
+from .engine import HedgeSynthesisEngine
+from .adapters import PriceFeedAdapter, OptionsVenueAdapter
+
+__all__ = [
+    "Asset", "MarketSignal", "RiskState", "HedgePlanDelta", "GraphOfContracts",
+    "HedgeSynthesisEngine", "PriceFeedAdapter", "OptionsVenueAdapter",
+]
diff --git a/src/idea91_ml_cv_hedge/adapters.py b/src/idea91_ml_cv_hedge/adapters.py
new file mode 100644
index 0000000..10b1542
--- /dev/null
+++ b/src/idea91_ml_cv_hedge/adapters.py
@@ -0,0 +1,47 @@
+from __future__ import annotations
+
+"""Toy adapters to bootstrap interoperability between venues and the core engine.
+
+- PriceFeedAdapter: simulates a price feed (e.g., equities, spot, etc.).
+- OptionsVenueAdapter: simulates an options venue with basic option data.
+
+These are deliberately simple and deterministic to serve as MVP building blocks.
+"""
+
+from typing import Dict, List
+from .dsl import Asset, MarketSignal
+
+
+class PriceFeedAdapter:
+    def __init__(self, assets: List[Asset]):
+        self.assets = assets
+        # deterministic seed values per asset
+        self.prices: Dict[str, float] = {a.symbol: 100.0 for a in assets}
+
+    def update(self) -> List[MarketSignal]:
+        # simple random walk with fixed seed for deterministic tests
+        signals: List[MarketSignal] = []
+        for a in self.assets:
+            price = self.prices[a.symbol] * 1.0005  # tiny drift
+            self.prices[a.symbol] = price
+            signals.append(MarketSignal(asset=a, signal_type="price", value=price))
+        return signals
+
+
+class OptionsVenueAdapter:
+    def __init__(self, underlying: Asset, strikes: List[float], maturities: List[float]):
+        self.underlying = underlying
+        self.strikes = strikes
+        self.maturities = maturities
+
+    def quote(self) -> List[MarketSignal]:
+        # toy: simple intrinsic value proxy for options across strikes
+        signals: List[MarketSignal] = []
+        for k in self.strikes:
+            delta = max(0.0,  self.underlying_asset_spot() - k)
+            signals.append(MarketSignal(asset=self.underlying, signal_type=f"option_delta_{k}", value=delta))
+        return signals
+
+    def underlying_asset_spot(self) -> float:
+        # deterministic spot via a fixed function of symbol hash
+        return float(sum(ord(c) for c in self.underlying.symbol)) % 200 + 50
diff --git a/src/idea91_ml_cv_hedge/delta_sync.py b/src/idea91_ml_cv_hedge/delta_sync.py
new file mode 100644
index 0000000..e5b764c
--- /dev/null
+++ b/src/idea91_ml_cv_hedge/delta_sync.py
@@ -0,0 +1,51 @@
+from __future__ import annotations
+
+import json
+import time
+from pathlib import Path
+from typing import List
+
+from .dsl import HedgePlanDelta
+
+
+class DeltaSync:
+    """Deterministic delta-sync with islanding replay capability.
+
+    Deltas are persisted to a log file in a deterministic, append-only manner.
+    Replays reconstruct the same sequence of delta applications on reconnect.
+    """
+
+    def __init__(self, log_dir: str = ".delta_logs"):  # hidden dir for replay logs
+        self.log_dir = Path(log_dir)
+        self.log_dir.mkdir(exist_ok=True)
+        self.log_file = self.log_dir / "delta_log.jsonl"
+
+    def log_delta(self, delta: HedgePlanDelta) -> None:
+        entry = {
+            "timestamp": delta.timestamp,
+            "contract_id": delta.contract_id,
+            "assets": [a.symbol for a in delta.assets],
+            "hedges": delta.hedges,
+            "author": delta.author,
+            "signature": delta.signature,
+        }
+        with self.log_file.open("a", encoding="utf-8") as f:
+            f.write(json.dumps(entry) + "\n")
+
+    def replay_all(self) -> List[ HedgePlanDelta ]:
+        deltas: List[HedgePlanDelta] = []
+        if not self.log_file.exists():
+            return deltas
+        with self.log_file.open("r", encoding="utf-8") as f:
+            for line in f:
+                obj = json.loads(line)
+                # Minimal reconstruction; actual assets would be resolved by registry in real system
+                delta = HedgePlanDelta(
+                    contract_id=obj.get("contract_id"),
+                    hedges=obj.get("hedges", {}),
+                    timestamp=obj.get("timestamp", time.time()),
+                    author=obj.get("author", "log"),
+                    assets=[ ]  # simplified; in test we don't rely on full asset objects here
+                )
+                deltas.append(delta)
+        return deltas
diff --git a/src/idea91_ml_cv_hedge/dsl.py b/src/idea91_ml_cv_hedge/dsl.py
new file mode 100644
index 0000000..ba4a0c8
--- /dev/null
+++ b/src/idea91_ml_cv_hedge/dsl.py
@@ -0,0 +1,58 @@
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+from typing import List, Dict, Optional
+import uuid
+import time
+
+
+@dataclass(frozen=True)
+class Asset:
+    symbol: str  # e.g., AAPL, SPY
+    asset_type: str  # e.g., 'equity', 'option'
+    locale: Optional[str] = None
+
+
+@dataclass(frozen=True)
+class MarketSignal:
+    asset: Asset
+    signal_type: str  # e.g., 'price', 'vol', 'delta'
+    value: float
+    timestamp: float = field(default_factory=lambda: time.time())
+
+
+@dataclass
+class RiskState:
+    asset: Asset
+    pnl_exposure: float  # expected short-term PnL exposure
+    delta_exposure: float  # net delta exposure
+    gamma_exposure: float  # net gamma exposure
+    volatility_shift: float  # forecasted vol shift
+
+
+@dataclass
+class HedgePlanDelta:
+    contract_id: str = field(default_factory=lambda: str(uuid.uuid4()))
+    assets: List[Asset] = field(default_factory=list)
+    hedges: Dict[str, float] = field(default_factory=dict)  # venue->notional_delta
+    objectives: Dict[str, float] = field(default_factory=dict)  # e.g., {'risk': 0.0}
+    timestamp: float = field(default_factory=lambda: time.time())
+    author: str = "system"
+    signature: Optional[str] = None
+
+
+class GraphOfContracts:
+    """Simple registry scaffold to-version adapters and data contracts."""
+
+    def __init__(self) -> None:
+        self.registry: Dict[str, Dict] = {}
+
+    def register(self, contract_name: str, schema: Dict) -> None:
+        self.registry[contract_name] = schema
+
+    def get(self, contract_name: str) -> Optional[Dict]:
+        return self.registry.get(contract_name)
+
+
+# Simple container to group DSL primitives for easy import
+__all__ = ["Asset", "MarketSignal", "RiskState", "HedgePlanDelta", "GraphOfContracts"]
diff --git a/src/idea91_ml_cv_hedge/engine.py b/src/idea91_ml_cv_hedge/engine.py
new file mode 100644
index 0000000..a2d65ca
--- /dev/null
+++ b/src/idea91_ml_cv_hedge/engine.py
@@ -0,0 +1,49 @@
+from __future__ import annotations
+
+from dataclasses import dataclass
+from typing import List, Dict
+import time
+
+from .dsl import Asset, MarketSignal, RiskState, HedgePlanDelta
+
+
+@dataclass
+class Budget:
+    risk_limit: float
+    liquidity_limit: float
+    latency_limit_ms: int
+
+
+class HedgeSynthesisEngine:
+    """A tiny, deterministic hedge synthesis engine.
+
+    This MVP uses a naive greedy approach: for each asset, compute a hedge delta per venue
+    aiming to reduce net delta exposure towards zero, constrained by budgets.
+    """
+
+    def __init__(self, budgets: Budget) -> None:
+        self.budgets = budgets
+
+    def assess_risk(self, risk_states: List[RiskState]) -> Dict[str, float]:
+        # aggregate risk metrics per asset (very simplified)
+        handoff: Dict[str, float] = {}
+        for rs in risk_states:
+            handoff[rs.asset.symbol] = rs.delta_exposure
+        return handoff
+
+    def synthesize(self, risk_states: List[RiskState], available_venues: List[str]) -> HedgePlanDelta:
+        # Very simple heuristic: allocate deltas inversely proportional to current delta exposure
+        hedge = HedgePlanDelta(
+            assets=[r.asset for r in risk_states],
+            hedges={},
+            timestamp=time.time(),
+            author="engine",
+        )
+        total_delta = sum(abs(r.delta_exposure) for r in risk_states) or 1.0
+        per_venue = 1.0 / len(available_venues) if available_venues else 0.0
+        for rs in risk_states:
+            # normalize, then assign to first venue as a toy example
+            alloc = (-rs.delta_exposure) / total_delta if total_delta else 0.0
+            if available_venues:
+                hedge.hedges[available_venues[0]] = hedge.hedges.get(available_venues[0], 0.0) + alloc * per_venue
+        return hedge
diff --git a/test.sh b/test.sh
new file mode 100644
index 0000000..31cc80b
--- /dev/null
+++ b/test.sh
@@ -0,0 +1,11 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+echo "Running tests..."
+export PYTHONPATH=src
+pytest -q
+
+echo "Building package..."
+python3 -m build
+
+echo "All tests passed and package built."
diff --git a/tests/test_core.py b/tests/test_core.py
new file mode 100644
index 0000000..84d48cd
--- /dev/null
+++ b/tests/test_core.py
@@ -0,0 +1,28 @@
+import math
+from idea91_ml_cv_hedge.dsl import Asset, MarketSignal, RiskState, HedgePlanDelta, GraphOfContracts
+from idea91_ml_cv_hedge.engine import HedgeSynthesisEngine, Budget
+from idea91_ml_cv_hedge.adapters import PriceFeedAdapter
+
+
+def test_dsl_asset_and_riskstate_roundtrip():
+    a = Asset(symbol="AAPL", asset_type="equity")
+    rs = RiskState(asset=a, pnl_exposure=1.0, delta_exposure=0.5, gamma_exposure=0.1, volatility_shift=0.2)
+    assert rs.asset == a
+    assert rs.delta_exposure == 0.5
+
+
+def test_hedge_plan_delta_basic():
+    a = Asset(symbol="AAPL", asset_type="equity")
+    rs = RiskState(asset=a, pnl_exposure=1.0, delta_exposure=0.5, gamma_exposure=0.1, volatility_shift=0.2)
+    delta = HedgePlanDelta(assets=[a], hedges={"Venue1": -0.5}, author="tester")
+    assert delta.assets[0] == a
+    assert delta.hedges["Venue1"] == -0.5
+
+
+def test_hedge_engine_basic_flow():
+    a = Asset(symbol="AAPL", asset_type="equity")
+    rs = RiskState(asset=a, pnl_exposure=1.0, delta_exposure=0.6, gamma_exposure=0.1, volatility_shift=0.2)
+    engine = HedgeSynthesisEngine(Budget(risk_limit=1.0, liquidity_limit=1.0, latency_limit_ms=100))
+    delta = engine.synthesize([rs], ["Venue1", "Venue2"])
+    assert isinstance(delta, HedgePlanDelta)
+    assert delta.assets[0] == a