Dynamic risk governance beyond static guards

[likidodefi]

OpenAlice's Guard Pipeline is a great foundation for per-trade safety checks — max position size and cooldowns cover the basics well.

One area that could take it further: regime-aware dynamic sizing. The current thresholds are static, which means the guards behave the same whether the market is trending, choppy, or in crisis.

A few ideas for extending the Guard Pipeline:

• Regime detection: classify market state (trending / mean-reverting / high-volatility) and adjust position limits accordingly. In a crisis regime, max size drops automatically.
• Macro-adaptive sizing: factor in cross-asset signals (rates, volatility indices, correlation shifts) to modulate how aggressive the guards allow the agent to be.
• Portfolio-level exposure: beyond per-trade guards, enforce total portfolio exposure caps across all active positions.
• Drawdown circuit breaker: pause all new entries when portfolio drawdown exceeds a threshold, regardless of individual trade signals.

I've been building a deterministic policy engine for autonomous trading agents that handles exactly this — happy to discuss approaches or potential integration points: https://github.com/likidodefi/riskstate-mcp