Single tool 'duckprobe_scan' has a clear, concise description (95 chars) and a simple input schema with one string parameter. However, the parameter lacks a description explaining what constitutes a valid 'target' (file path format, warehouse connection string syntax, supported file types). No output schema is documented, the tool returns JSON but the structure is not specified. The tool name uses verb_noun pattern correctly but the implementation is minimal. No error handling guidance, no pagination, no validation constraints documented.

• Add a detailed parameter description for 'target': 'Path to a local CSV, Parquet, or JSON file; or a DuckDB connection string (e.g., duckdb:///path/to/db.duckdb or postgresql://user:pass@host/db). Supported file types: .csv, .parquet, .json, .jsonl.'
• Document the output schema in the tool description: 'Returns a JSON object with keys: findings (array of {id, severity, message, location}), score (integer count of issues), and summary (string).'
• Expand tool description to explain what checks are performed: 'Runs bundled data-quality checks including row count, null detection, type validation, and anomaly detection. No configuration required, uses sensible defaults.'
• Add error handling guidance: 'If the file is not found or unsupported, returns an error with the path and supported formats. If a warehouse connection fails, returns the connection error and suggests checking credentials.'
• Specify constraints in parameter description: 'Must be a valid file path (max 1024 chars) or connection string. Relative paths are resolved from the current working directory.'