ai-trading-skill

Explainable Trading Intelligence Skill

This skill equips any agent with the ability to run market analysis, generate explainable trade rationales, backtest strategies, evaluate risk, and manage execution on the Hyperliquid exchange.

Prerequisite Setup

Before running any script, activate the workspace environment:

cd <project-root>
source .venv/bin/activate

User must configure private keys in .env for active trades. See .env.example for required fields.

Core Operations

1. Run Market Analysis with Trade Rationale

The primary output of this framework. Generates a structured trade rationale covering trend, volatility, momentum, key levels, sizing, and invalidity conditions.

# Default analysis (BTC, moderate risk profile)
python3 analyze.py

# Custom asset, interval, and risk profile
python3 analyze.py --symbol ETH --interval 5m --risk-profile conservative

# Adjust per-trade risk percentage
python3 analyze.py --symbol SOL --account-risk-pct 0.5

Available risk profiles: conservative, moderate, aggressive

The output includes four sections:

• •Strategy scoring matrix (all 5 layers)
• •Trade Rationale Engine output (structured breakdown)
• •Risk Assessment (profile-adjusted sizing and warnings)
• •Institutional Context (macro positioning data)

2. Run Automated Tests

Verify strategy calculations, rationale engine, and risk layer:

python3 -m unittest tests/test_strategies.py

3. Run Historical Backtests

Walk-forward simulation with no lookahead bias:

python3 backtest.py
python3 backtest.py --days 60 --confidence 75 --min-agree 4

4. Evaluate Position Sizing Models

Calculate compounding returns and drawdowns from backtest results:

python3 pnl_calc.py

5. View Signals Breakdown

Inspect per-trade strategy agreement matrix:

python3 show_signals.py

6. Run Execution Orchestrator

Full execution loop with safety gates, risk circuit breakers, and LLM audit:

# One-shot dry tick
python3 main.py

# Continuous polling loop
python3 main.py --loop

Configuration

Edit config.yaml to adjust:

• •agree_threshold: Minimum confidence for a strategy to count as agreeing (default: 50)
• •min_agree: Minimum strategies required for consensus (default: 4)
• •symbol: Asset to analyze (default: BTC)
• •interval: Candle timeframe (default: 15m)
• •Individual strategy parameters under the strategies section