The 'scan' tool has a minimal definition with a short description (71 chars) and a single parameter. The description lacks context on when to use the tool, what it returns, or prerequisites. The parameter 'target' has a basic description but no type constraints, format guidance, or examples of valid input. No output schema is documented. The tool definition is visible in cryptotrace/mcp_server.py but lacks the depth required for LLM-driven selection and error recovery.

• Expand tool description to 100-200 chars: 'Screen a blockchain transaction list for OFAC sanctions exposure, taint propagation, address clustering, and peeling-chain laundering patterns. Accepts a file path or raw JSON/JSONL string. Returns a structured report with findings, severity, and actor tags.'
• Document the output schema explicitly in the tool definition or docstring. Include fields like 'findings' (array), 'clusters' (array), 'taint_score' (number), 'sanctioned_addresses' (array).
• Add parameter constraints: 'target (string): File path to a JSON/JSONL transaction list, or raw JSON/JSONL text. If a file path, must be readable and valid UTF-8. If raw JSON, must be valid JSON or JSONL format.'
• Rename tool from 'scan' to 'screen_transactions' or 'analyze_blockchain_transactions' to clarify the action and domain.
• Add error handling documentation: 'If the file cannot be read, the tool treats the input as raw JSON. If JSON parsing fails, returns an error with the parse failure reason and line number.'