Skip to main content
Back to challenges
Challenge

Build a Credit Risk Early-Warning Agent with Claude Agents SDK and Daytona

Rising UK credit card defaults require automated early-warning risk monitoring to predict defaults before charge-offs occur. Build a sandboxed credit risk triage agent using Claude Agents SDK and Daytona that analyzes transaction velocity and delinquency indicators.

CybersecurityHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Develop an isolated credit monitoring agent running inside Daytona containers using Claude Agents SDK to flag emerging credit default indicators.

Delivery guide

How work is evaluated

Evaluation

Evaluates default prediction recall and risk score calibration against credit behavior datasets.

Datasets

Challenge datasets

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

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

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

test_high_utilization_alert

Ensure high default probability predicted for maxed-out card sequence

Binary check

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

Dimension 2

auc_roc

Area under the receiver operating characteristic curve • target: 0.9 • range: 0.8-1

Binary check

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

Learning goals

What you should walk away with

  • Set up isolated execution sandboxes using Daytona for secure risk scoring script execution.

  • Orchestrate Claude Agents SDK with extended thinking to evaluate revolving debt signals.

  • Generate automated mitigation actions such as line reduction or proactive outreach.

Resources and assets

Reference links and supporting material

Dataset notes

Synthetic time-series credit card account histories including delinquency transitions.

Challenge setup

Evaluates default prediction recall and risk score calibration against credit behavior datasets.

Challenge input

JSON timeseries of user credit card transactions and balance history

Claude Agents SDK

Agent framework assignment requirement

Daytona

Isolated sandbox assignment requirement

Evaluated output

JSON containing default_probability_90d and recommended_action

Evaluation checks
  • Ensure high default probability predicted for maxed-out card sequence
  • Area under the receiver operating characteristic curve • target: 0.9 • range: 0.8-1
Configured evaluation evidence
  • Benchmark: Credit Risk Benchmark v1
  • Auc Roc target: 0.9
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.

View technical recipe

Configured tools

Environment
  • Daytona · Required
Action Space
  • Daytona · Required
  • Claude Code · Optional
Observation
  • Claude Code · Optional
Policy Serving
  • Claude 4.1 Opus · Optional

Evaluation contract

  • test_high_utilization_alert · Weight 1
  • auc_roc · Weight 1

Metrics

  • auc_roc · Target 0.9

Benchmarks

  • Credit Risk Benchmark v1
  • security

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON timeseries of user credit card transactions and balance history

    Output: JSON containing default_probability_90d and recommended_action

    Execution harness: Python

Evaluation tools

  • Claude Agents SDK · agent_framework
  • Daytona · agent_environment

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Build a Credit Risk Early-Warning Agent with Claude Agents SDK and Daytona