Server has 20 tools with mostly complete schemas and descriptions in Spanish. Naming follows verb_noun pattern (load_project, save_project, list_layers). Descriptions average ~80 chars, meeting baseline minimums. However, critical gaps exist: (1) No output schemas documented for any tool, LLMs cannot predict return structure; (2) Parameter descriptions are minimal (often 1-2 words); (3) No error handling guidance, tools return {ok: bool, error: str} but don't guide recovery; (4) execute_pyqgis is an unguarded escape hatch with IRREVERSIBLE risk; (5) Missing enum constraints on classification_type, orientation, field_type despite fixed valid values. Tools like calculate_field and get_attribute_table show good internal validation but don't expose it in descriptions. Per-tool average: 62/100.

• Document output schema for every tool. E.g., list_layers should specify: {ok: bool, layers: [{id: string, name: string, type: string, crs: string, feature_count: int}], total: int}. This enables LLM to chain tools correctly.
• Add enum constraints to fixed-value parameters: set_layer_style.classification_type → enum: ['categorized', 'graduated']; calculate_field.field_type → enum: ['int', 'double', 'string', 'bool']; create_print_layout.orientation → enum: ['landscape', 'portrait'].
• Expand parameter descriptions to 50-150 chars with format/constraint hints. E.g., load_project.path: 'Path to QGIS project file (.qgs or .qgz). Absolute or relative to working directory. Must exist and be readable.'
• Add error recovery guidance to tool descriptions. E.g., load_project: 'If file not found, call list_layers first to verify project is loaded, or check file path with a file browser tool.'
• Restrict execute_pyqgis: require explicit user confirmation before execution, or remove it entirely and expose specific PyQGIS operations as dedicated tools (e.g., get_feature_geometry, apply_spatial_filter).