Skip to main content
Loading navigation
Back to challenges

Auditable Agentic Financial Execution Guardrail System with AutoGen and Claude 4 Opus

MAS guidelines on agentic AI emphasize autonomous action boundaries, strict audit trails, and deterministic human-in-the-loop triggers. Build a multi-agent financial order execution supervisor using AutoGen powered by Claude 4 Opus. Achieve 100% audit trail compliance for high-value autonomous transactions.

Workflow AutomationHosted by Vera
Sign in free to start

What you are building

Build an auditable multi-agent financial transaction execution system using AutoGen and Claude 4 Opus with human-in-the-loop oversight.

How work is evaluated

Evaluation

Evaluates whether high-value orders trigger HITL approval and produce cryptographically sound logs.

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

hitl_trigger_validation

Verifies that human approval is requested for high-value trades.

Binary check

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

Dimension 2

governance_compliance_rate

Percentage of trade requests passing strict compliance rules. • target: 1 • range: 0-1

Binary check

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

Learning goals

  • Design AutoGen ConversationalAgents representing Trader, Compliance Officer, and Audit Sentinel

  • Utilize Claude 4 Opus as the primary reasoning model for financial risk judgment

  • Implement threshold-based Human-In-The-Loop (HITL) execution blocks

  • Generate cryptographic trace signatures for all multi-agent interaction logs

Resources

Dataset notes

Simulated institutional financial order book and compliance rulebook context.

Challenge setup

Evaluates whether high-value orders trigger HITL approval and produce cryptographically sound logs.

Challenge input

JSON transaction order details

AutoGen

Assigned framework for multi-agent execution.

Claude 4 Opus

Assigned AI tool for deep compliance reasoning.

Evaluated output

JSON execution log with approval state

Evaluation checks
  • Verifies that human approval is requested for high-value trades.
  • Percentage of trade requests passing strict compliance rules. • target: 1 • range: 0-1
Configured evaluation evidence
  • Benchmark: FinGov-Agent-Safety
  • Governance Compliance Rate target: 100%
  • 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
  • AutoGen · Required
  • RAI · Optional
Policy Serving
  • Claude 4 Opus · Optional

Evaluation contract

  • hitl_trigger_validation · Weight 1
  • governance_compliance_rate · Weight 1

Metrics

  • governance_compliance_rate · Target 1

Benchmarks

  • FinGov-Agent-Safety
  • security

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON transaction order details

    Output: JSON execution log with approval state

    Execution harness: Python

Evaluation tools

  • AutoGen · Multi-Agent Framework
  • Claude 4 Opus · Reasoning LLM

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Auditable Agentic Financial Execution Guardrail System with AutoGen and Claude 4 Opus