Skip to main content
Loading navigation
Back to challenges

Claude Agents SDK Data Centre Security Bill Compliance Auditor

New data sovereignty and security bills mandate local encryption key handling and strict cross-border transfer logs for data hubs. Build an automated cloud architecture auditor using Claude Agents SDK to continuously inspect infrastructure configurations.

Business OperationsHosted by Vera
Sign in free to start

What you are building

Develop an automated data sovereign cloud compliance auditor using Claude Agents SDK.

How work is evaluated

Evaluation

Evaluates cloud security configuration audit reports.

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

violation_detected

Checks correct flagging of non-compliant region.

Binary check

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

Dimension 2

audit_recall

Recall metric on misconfigurations • 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

  • Deploy Claude Agents SDK with vision and code inspection capabilities

  • Audit Terraform / AWS IAM security policies against Data Centre Security Bill rules

  • Automate generation of remediation scripts for identified non-compliant buckets

Resources

Dataset notes

Cloud security infrastructure state files and regulatory policy specs.

Challenge setup

Evaluates cloud security configuration audit reports.

Challenge input

JSON cloud config summary

Claude Agents SDK

Mandatory framework requirement

Evaluated output

JSON audit result and patch code

Evaluation checks
  • Checks correct flagging of non-compliant region.
  • Recall metric on misconfigurations • target: 1 • range: 0-1
Configured evaluation evidence
  • Benchmark: Cloud-Security-Bench
  • Audit Recall target: 1
  • 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
  • Ray Tune · Required
  • AutoML (H2O) · Optional
  • Bland AI · Optional

Evaluation contract

  • violation_detected · Weight 1
  • audit_recall · Weight 1

Metrics

  • audit_recall · Target 1

Benchmarks

  • Cloud-Security-Bench
  • security

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON cloud config summary

    Output: JSON audit result and patch code

    Execution harness: Python

Evaluation tools

  • Claude Agents SDK · Agent Framework

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Claude Agents SDK Data Centre Security Bill Compliance Auditor