Orderflow Analysis Skill
Detects institutional trading patterns from Level 2 market data and trade executions.
Capabilities
This skill enables the agent to:
- Analyze L2 orderbook depth for bid/ask walls
- Detect absorption patterns (hidden liquidity)
- Detect exhaustion at support/resistance
- Identify imbalance sweeps
- 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 snapshot3. Run Detection Algorithms
When sufficient data is collected:
Call: analyze_footprint(symbol: str, window_seconds: int)
Returns: List[FootprintSignal] with pattern type, direction, confidence4. Interpret Signals
| Signal Type | Description | Suggested Action |
|---|---|---|
| ABSORPTION | Heavy volume absorbed without price movement | Fade the volume direction |
| EXHAUSTION | Declining volume at S/R | Prepare for reversal |
| IMBALANCE | 3:1+ buy/sell ratio | Follow imbalance direction |
| SWEEP | Multiple levels cleared rapidly | Momentum follow |
5. Forward to Confirmation Mesh
All signals must pass through confirmation mesh before execution:
Call: validate_signal(signal: FootprintSignal, quantity: float) -> ConfirmationResultSafety 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)