Single tool 'lineagemap_scan' has a basic description (58 chars) but critically lacks parameter descriptions. The input schema defines 'target' as a string with a description, but the tool description itself is vague about what 'column-level lineage' means, when to use it, or what the JSON output structure contains. No output schema is documented. The tool name uses a verb+noun pattern (good), but the description does not explain prerequisites, expected file formats, or error conditions. No error handling guidance is provided.

• Add a detailed parameter description for 'target': 'Path to a directory or SQL file to scan. Accepts absolute or relative paths. Recursively searches for .sql and dbt model files. Returns empty findings if no SQL files are found.'
• Expand tool description to 150 - 250 chars explaining: (1) what it does (extracts column-level lineage from SQL/dbt), (2) when to use it (when analyzing data pipeline dependencies), (3) what it returns (JSON with findings array containing lineage relationships).
• Document the output schema explicitly: 'Returns {tool: string, findings: [{id: string, severity: string, location: string, ...}], score: number}' so the LLM knows what fields to expect.
• Add error handling guidance: 'If target is not found, returns empty findings. If target contains no SQL files, returns {findings: [], score: 0}. If parsing fails, returns error with file path and line number.'
• Rename tool to 'scan_sql_lineage' or 'analyze_column_lineage' to make the action and domain explicit without requiring the LLM to infer from description.