Skip to main content
Loading navigation
Back to challenges

Autonomous FX Risk Mitigation Agent with OpenAI Agents SDK and Claude 4 Sonnet

With institutions rapidly scaling cross-border liquidity operations, institutional FX risk requires proactive hedging strategies that execute in fractions of a second when exposure limits are breached. Autonomous agents can actively monitor rate differentials and trigger automated forward contracts or swap orders. In this challenge, you will implement an autonomous FX hedging agent using the OpenAI Agents SDK powered by Claude 4 Sonnet. The agent will monitor real-time multi-currency exposure and autonomously trigger hedge executions when volatility indices cross predefined risk bands.

CybersecurityHosted by Vera
Sign in free to start

What you are building

Build an autonomous FX risk mitigation agent using OpenAI Agents SDK and Claude 4 Sonnet to hedge dynamic multi-currency positions.

How work is evaluated

Evaluation

Evaluates portfolio value protection under simulated currency market shocks.

Datasets

Public datasets, plus the private datasets you have access to.

How submissions are scored

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

exposure_limit_enforced

Verifies unhedged risk is driven below target max exposure

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

sharpe_ratio_improvement

Relative improvement in portfolio Sharpe ratio post-hedging • target: 1.8 • range: 0-5

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

  • Set up OpenAI Agents SDK agent definitions with custom dynamic function tools

  • Integrate Claude 4 Sonnet via API adapter endpoints for advanced mathematical reasoning

  • Enforce maximum downside loss limits through strict tool guardrails

  • Build real-time position audit traces for treasury risk officer evaluation

Resources

Dataset notes

Historical foreign exchange spot and forward curves under periods of elevated macroeconomic volatility.

Challenge setup

Evaluates portfolio value protection under simulated currency market shocks.

Challenge input

JSON portfolio snapshot and FX rate shock matrix

OpenAI Agents SDK

Official framework for multi-turn autonomous agent loops.

Claude 4 Sonnet

Advanced model offering superior mathematical reasoning and precision.

Evaluated output

JSON containing execution_orders and unhedged_risk_ratio

Evaluation checks
  • Verifies unhedged risk is driven below target max exposure
  • Relative improvement in portfolio Sharpe ratio post-hedging • target: 1.8 • range: 0-5
Configured evaluation evidence
  • Benchmark: FinRisk-Hedging-Bench
  • Sharpe Ratio Improvement target: 1.8
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Shows the first task, up to three tools, and up to three evidence items.

View technical recipe

Configured tools

Action Space
  • OpenAI · Required
  • Claude Code · Optional
Observation
  • Claude Code · Optional
Policy Serving
  • Claude 4 Sonnet · Optional

Evaluation contract

  • exposure_limit_enforced · Weight 1
  • sharpe_ratio_improvement · Weight 1

Metrics

  • sharpe_ratio_improvement · Target 1.8

Benchmarks

  • FinRisk-Hedging-Bench
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON portfolio snapshot and FX rate shock matrix

    Output: JSON containing execution_orders and unhedged_risk_ratio

    Execution harness: Python

Evaluation tools

  • OpenAI Agents SDK · Agent Framework
  • Claude 4 Sonnet · Language Model

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Autonomous FX Risk Mitigation Agent with OpenAI Agents SDK and Claude 4 Sonnet