alpha-trading-system

Files (5)

• CONTRACT-REFERENCE.md
• DOCUMENTATION-WORKFLOW.md
• PATTERNS-REFERENCE.md
• SESSION-WORKFLOW.md
• SKILL.md

SKILL.md (first lines)

# Alpha AI Trading System Skill

Comprehensive skill for working on the `alpha_data_scraper_ai` project. This hub points to detailed workflows and reference materials.

---

## Quick Start (5 minutes)

### Common Tasks

**Working on signals?**

- See: [SIGNAL-GENERATION.md](./SIGNAL-GENERATION.md)
- Files: `signal_generator.py`, `multi_timeframe.py`, `claude_ai.py`

**Working on data/MT5?**

- See: [DATA-FETCHING.md](./DATA-FETCHING.md)
- Files: `mt5_fetcher.py`, `mt5_trader.py`, `metrics_server.py`

**Working on risk management?**

- See: [RISK-GATES.md](./RISK-GATES.md)
- Files: `risk/risk_manager.py`, `calculator.py`

**Working on ML/LSTM?**

- See: [LSTM-PIPELINE.md](./LSTM-PIPELINE.md)
- Files: `lstm_model.py`, `ai/lstm_model.py`, `indicators.py`

**Writing tests?**

- See: [TESTING-PATTERNS.md](./TESTING-PATTERNS.md)
- Files: `tests/conftest.py`, `tests/test_*.py`

**Setting up documentation?**

- See: [DOCUMENTATION-WORKFLOW.md](./DOCUMENTATION-WORKFLOW.md)

---

## MCP Tool Integration

Before any work, establish context with MCP:

```python
# Python example (or use terminal commands)
from mcp_client import MCPClient

client = MCPClient()

# 1. Identify context
context = client.identify_context(file_path="./ai_engine.py")
print(context)  # { "project": "alpha_data_scraper_ai", "context": "backend" }

# 2. Get current focus
focus = client.get_current_focus()

# 3. Load guidelines
guidelines = client.get_merged_guidelines(context="backend")

# 4. Start session if needed
if not focus:
    client.start_session({
        "context": "backend",
        "current_focus": "signal generation",
        "goal": "Implement unified signal aggregation"
    })

# 5. Do your work...

# 6. Create checkpoint when changing focus
client.create_checkpoint({
    "summary": "Implemented signal clamping [33-85]",
    "next_focus": "Claude validation"
})
```

**Critical**: Always use relative paths: `./path/file.py` not `/absolute/path`

---

## Architecture Map

The system follows a **Progressive Disclosure** pattern:

### Layer 1: Minimal Core Context

Start here when first looking at a file:

- **Entry Point**: `main.py` — CLI interface
- **Orchestrator**: `ai_engine.py` — Signal pipeline coordination
- **Config**: `config.json` — Runtime parameters

### Layer 2: Domain-Specific Modules

Dig deeper based on what you're working on:

**Data Layer**: `mt5_fetcher.py`, `mt5_trader.py`
**Analysis Layer**: `indicators.py`, `signal_generator.py`, `lstm_model.py`
**Fusion Layer**: `multi_timeframe.py`, `news_sentiment.py`
**AI Layer**: `claude_ai.py`, `ai_engine.py`
**Risk Layer**: `risk/risk_manager.py`
**Output Layer**: `mt5_trader.py`, `metrics_server.py`

### Layer 3: Advanced References

Pull in as needed:

- **Testing**: See TESTING-PATTERNS.md
- **Patterns**: See PATTERNS-REFERENCE.md
- **Contracts**: See CONTRACT-REFERENCE.md
- **Session Workflow**: See SESSION-WORKFLOW.md

---

## Key Concepts

### Signal Confidence

From Melly-999/alpha_data_scraper_ai, Apache-2.0.

More from Melly-999/alpha_data_scraper_ai