orderflow-analysis（订单流分析）

Orderflow Analysis Skill

Detects institutional trading patterns from Level 2 market data and trade executions.

Capabilities

This skill enables the agent to:

1. Analyze L2 orderbook depth for bid/ask walls
2. Detect absorption patterns (hidden liquidity)
3. Detect exhaustion at support/resistance
4. Identify imbalance sweeps
5. Generate trade signals with confidence levels

Prerequisites

• Active L2 data connection (Alpaca Pro or Polygon)
• Trading symbols configured in watchlist

Procedural Steps

1. Connect to L2 Data Stream

Use the trading-orderflow MCP server to establish WebSocket connection.
Call: connect_l2_stream(symbol: str, provider: "alpaca" | "polygon")

2. Monitor Orderbook State

Track bid/ask walls and imbalance ratios.
Call: get_orderbook_state(symbol: str) -> returns current book snapshot

3. Run Detection Algorithms

When sufficient data is collected:

Call: analyze_footprint(symbol: str, window_seconds: int)
Returns: List[FootprintSignal] with pattern type, direction, confidence

4. Interpret Signals

5. Forward to Confirmation Mesh

All signals must pass through confirmation mesh before execution:

Call: validate_signal(signal: FootprintSignal, quantity: float) -> ConfirmationResult

Safety Guardrails

• Never execute trades based on LOW confidence signals
• Require L2 liquidity verification before market orders
• All executions must go through confirmation_mesh validation
• Circuit breakers halt trading after consecutive failures

Example Workflow

# Agent detects high-confidence absorption
signal = await analyze_footprint("AAPL", window_seconds=60)

if signal.signal_type == "ABSORPTION" and signal.confidence == "HIGH":
    # Validate before execution
    result = await validate_signal(signal, quantity=100)

    if result.approved:
        # Proceed to execute-trade skill
        await execute_confirmed_trade(result)