Claude Agents SDK Virtual Asset Compliance Auditor
Implement an extended-thinking agent using Claude Agents SDK to audit Virtual Asset Service Providers (VASPs) for travel rule compliance and blockchain transaction risk. Reach 94% alignment with MAS regulatory enforcement guidelines.
What you are building
Build an extended-reasoning compliance auditor using Claude Agents SDK to enforce Travel Rule and VASP guidelines.
How work is evaluated
Evaluates Travel Rule detection accuracy and completeness of generated VASP compliance audits.
Datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
Trigger Threshold Test
Ensures Travel Rule triggers for transfers over 1500 SGD
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
regulatory_alignment
Alignment with MAS enforcement rules • target: 0.94 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Learning goals
Leverage Anthropic's Claude Agents SDK with extended thinking for deep regulatory analysis
Implement tools for blockchain transaction graph lookup and VASP registry checks
Verify Travel Rule payload completeness (originator and beneficiary PII)
Generate regulator-ready audit reports with risk recommendations
Resources
Synthetic Virtual Asset transaction logs including mock Travel Rule protocol messages.
Challenge setup
Evaluates Travel Rule detection accuracy and completeness of generated VASP compliance audits.
Challenge input
JSON containing transfer payload, asset type, amount_sgd, and originator_data
Claude Agents SDK
Extended thinking capabilities for complex crypto regulatory analysis
Evaluated output
JSON containing travel_rule_triggered, compliant, and regulatory_flag
- Ensures Travel Rule triggers for transfers over 1500 SGD
- Alignment with MAS enforcement rules • target: 0.94 • range: 0-1
- Benchmark: CryptoComplianceBench
- Regulatory Alignment target: 0.94
- 1 public reference case
- 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
- n8n · Required
- Mindset AI · Optional
- Vercel AI SDK · Optional
- Vercel AI SDK · Optional
Evaluation contract
- Trigger Threshold Test · Weight 1
- regulatory_alignment · Weight 1
Metrics
- regulatory_alignment · Target 0.94
Benchmarks
- CryptoComplianceBench
- security
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing transfer payload, asset type, amount_sgd, and originator_data
Output: JSON containing travel_rule_triggered, compliant, and regulatory_flag
Execution harness: Python
Evaluation tools
- Claude Agents SDK · Agent Framework
Recipe freeze marker
No recipe freeze time is recorded.