ai agent risk governor

The problem

Autonomous trading agents can cause catastrophic financial loss if they execute unauthorized actions, exceed leverage limits, or encounter runaway bugs. Relying on an agent to manage its own risk is insufficient, as logic failures can bypass internal checks.

What it does

• Enforces hard, external limits on exposure, loss, leverage, and asset eligibility.
• Implements a fail-closed mechanism that blocks all actions if risk data is stale or unavailable.
• Triggers a hard kill switch requiring human intervention after any limit breach.
• Maintains an immutable, append-only audit trail of every ALLOW and BLOCK decision.
• Validates every proposed action against account state via a pre-flight check middleware.

Frameworks & tools

Language-agnostic logic designed for integration with Python, Node.js, or Go trading bots. Compatible with REST/WebSocket exchange APIs and YAML-based configuration.

Why this beats prompting it yourself

Generic prompts often allow agents to "hallucinate" ways around risk constraints or ignore edge cases in volatile markets. This skill provides a deterministic governance framework that treats risk as a separate, non-negotiable layer the agent cannot modify or influence.

Use cases

• Preventing runaway losses in autonomous crypto or forex trading bots.
• Restricting agents to specific asset allowlists and liquidity thresholds.
• Enforcing strict trading windows and mandatory cooldown periods after losses.
• Capping total account leverage and position concentration per sector.

Known limitations

Requires an external environment to host the governor logic so the agent cannot bypass it. Does not include direct exchange connectivity adapters.