Crypto Sanctions & Regulatory Compliance Agent with Claude Agents SDK
Cryptocurrency platforms face strict reporting mandates under proposed SEC regulations for detecting illegal cross-border token transfers. Build a compliance audit agent using Claude Agents SDK and DeepSeek R1 to analyze wallet transactions against OFAC sanctions lists, achieving 98% detection precision on illicit flows.
What you are building
The core problem, expected build, and operating context for this challenge.
Construct an automated crypto transaction compliance scanner using Claude Agents SDK and DeepSeek R1 reasoning to evaluate multi-chain wallet interactions against regulatory sanctions databases.
How work is evaluated
Evaluates the agent's ability to identify sanctioned crypto entities and output compliant audit logs.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
Exact SDN Identification
Ensure 100% of known SDN address transactions are correctly flagged.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
precision
Precision of illicit wallet flagging • target: 0.98 • range: 0.95-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Implement deep reasoning workflows using Claude Agents SDK and DeepSeek R1 for sanctions evasion detection
Design deterministic entity resolution routines connecting blockchain wallet addresses to legal entities
Automate Suspicious Activity Report (SAR) payload generation conforming to SEC guidelines
Reference links and supporting material
Synthetic multi-chain crypto transaction ledger containing 10,000 transactions with injected OFAC SDN illicit wallet paths.
Challenge setup
Evaluates the agent's ability to identify sanctioned crypto entities and output compliant audit logs.
Challenge input
JSON transaction log containing wallet addresses, amounts, and hop counts
Claude Agents SDK
Required primary agent framework for Anthropic models.
DeepSeek R1
Deep reasoning model for multi-hop graph analysis.
Evaluated output
JSON evaluation report with risk scores, flagged addresses, and justification string
- Ensure 100% of known SDN address transactions are correctly flagged.
- Precision of illicit wallet flagging • target: 0.98 • range: 0.95-1
- Benchmark: FinAML-Bench
- Precision target: 0.98
- 1 public reference case
- 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
- Claude Code · Required
- DeepSeek · Optional
- Claude Code · Required
- Claude 4.1 Opus · Optional
Evaluation contract
- Exact SDN Identification · Weight 1
- precision · Weight 1
Metrics
- precision · Target 0.98
Benchmarks
- FinAML-Bench
- security
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON transaction log containing wallet addresses, amounts, and hop counts
Output: JSON evaluation report with risk scores, flagged addresses, and justification string
Execution harness: Python
Evaluation tools
- Claude Agents SDK · Agent Orchestration
- DeepSeek R1 · Reasoning Engine
Recipe freeze marker
No recipe freeze time is recorded.