MCP server providing interactive tools for the Bilbo Frontend event with color picker, language statistics visualization, and survey results analysis
The server registers 4 tools with some schema and description coverage, but exhibits significant gaps across naming clarity, parameter descriptions, and output schema documentation. Tools use the Apps extension (registerAppTool, registerAppResource) and leverage Zod for schema validation, which is positive. However, parameter descriptions are inconsistently applied, output schemas are partially documented in registerAppTool calls but not exposed to the LLM in inputSchema format, and error handling is minimal. Naming is acceptable but not action-verb-forward. The average of per-tool scores (65, 60, 55, 52) yields 58.
Get information about the attendees
Opens an interactive color picker UI and returns the selected color.
Muestra un gráfico con el porcentaje de cada lenguaje de programación usado por los asistentes de Bilbo Frontend.
Muestra los resultados de la encuesta de Bilbo Frontend con gráficos de lenguajes, fuentes y niveles de experiencia. Si no se proporcionan respuestas, lee el archivo CSV por defecto.
Output schemas are defined in registerAppTool (outputSchema: { color: ... }) but are NOT visible in the exposed JSON Schema to the LLM, they exist as internal TypeScript objects, not as discoverable schema. LLMs cannot plan downstream tool calls without documented return types.
get_attendees has an empty inputSchema ({}). While this may be intentional (no parameters required), the tool description does not explicitly state what criteria it uses to fetch attendees or what the returned structure is. Output shape is undocumented.
show_language_stats and show_survey_results have descriptions in Spanish only ('Muestra un gráfico...') without English equivalents. LLMs operating in English may struggle with intent inference, and non-Spanish-fluent operators cannot review tool purpose.
Inferred effective spec: 2025-06-18+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 63 | 2025-06-18+ | v2 |
| 2026-03-09 | F | 42 | - | v1 |
Parameters 'responses' in show_survey_results is marked optional but the description says 'si no se proporciona lee el CSV' (if not provided, reads the CSV). This conditional logic is not formalized, no enum or explicit fallback guidance in the description itself.
No error handling or recovery guidance. Tools return structured content inline but do not specify what happens on failure (file not found, malformed CSV, invalid input). LLMs have no guidance on retry or alternatives.
Tool naming does not consistently use action verbs. 'show_language_stats' and 'show_survey_results' use 'show' (which may imply UI-only side effects), not 'get_' or 'list_'. 'pick_color' uses 'pick' rather than standard 'select_' or 'get_'. Inconsistency increases LLM cognitive load.