SKILL.md
z.ai Expert Team Integration
z.ai provides a team of domain experts accessible through the zread MCP server, working alongside Google Antigravity IDE.
Configuration ✅
Status: Configured and ready to use
Environment Variable (Windows PowerShell)
# Session only
$env:Z_AI_API_KEY = "your-z-ai-key"
# Permanent (User-level)
[Environment]::SetEnvironmentVariable("Z_AI_API_KEY", "your-key", "User")MCP Server Config
Located in .antigravity/mcp-config.json:
{
"zread": {
"type": "streamableHttp",
"url": "https://api.z.ai/api/mcp/zread/mcp",
"headers": {
"Authorization": "${Z_AI_API_KEY}"
}
}
}Available Experts
| Expert | Domain | Best For |
|---|---|---|
| 💰 Finance | Markets, Trading | Strategy, risk management, orderflow patterns |
| ⚙️ Engineering | Architecture, Code | System design, performance, debugging |
| 📊 Data Science | ML, Analytics | Signal processing, pattern detection, models |
| 🔐 Security | Compliance, Safety | API security, secrets, audit logging |
When to Consult z.ai
High-Value Scenarios
| Scenario | Expert | Example Query |
|---|---|---|
| Trading logic | Finance | "What orderflow patterns indicate institutional accumulation?" |
| System design | Engineering | "How should I structure WebSocket reconnection for HFT?" |
| ML features | Data Science | "What features best predict short-term price movements?" |
| Security review | Security | "What's the best approach for rotating API keys in production?" |
Decision Flow
graph LR
A[Question] --> B{Type?}
B -->|Trading/Market| C[z.ai Finance]
B -->|Code/Architecture| D[z.ai Engineering]
B -->|ML/Data| E[z.ai Data Science]
B -->|Security/Compliance| F[z.ai Security]
C --> G[Expert Analysis]
D --> G
E --> G
F --> G
G --> H[Antigravity Implements]Collaboration Pattern
z.ai + Antigravity Workflow
- Identify domain → Choose appropriate z.ai expert
- Ask specific question → Include context and constraints
- Receive expert analysis → z.ai provides recommendations
- Implement solution → Antigravity writes the code
- Validate → Test and verify
Best Practices
- Be Specific: Include platform context (latency, scale, market)
- Provide Constraints: Mention requirements and limitations
- Iterate: Refine approach based on expert feedback
- Document: Record recommendations for future reference
Example Usage
Consulting Finance Expert
"I'm building an orderflow system for Indian equities. What footprint
patterns should I detect to identify institutional activity, given
the 5% circuit limit and market hours of NSE?"Consulting Engineering Expert
"My WebSocket connection to Alpaca drops every few hours. What's the
best reconnection strategy for a trading system where I can't miss
more than 1 second of L2 data?"Guardrails
[!CAUTION]
- z.ai provides expertise, not financial advice
- Always validate recommendations against your requirements
- Test thoroughly in paper trading before live deployment
- Expert input should inform, not replace, your judgment