A Flask-based HTTP server that manages shared context data between multiple models via REST API endpoints
This server has critical definition quality gaps. Both tools lack proper parameter descriptions, the 'context' parameter in update_context has only a generic object description with no guidance on expected structure, format, or constraints. get_context has an empty input schema with no documentation. Descriptions are minimal (under 50 chars) and do not explain WHEN to use each tool, WHAT it returns, or error handling. No output schemas are documented. The server appears to be a bare Flask HTTP endpoint without MCP protocol integration, tool definitions are inferred from HTTP routes, not explicitly registered via MCP. This violates the 'tool registration' requirement and caps per-tool scores at 50 maximum.
Retrieves the current global context data
Updates the global context data with JSON payload from POST request
No explicit MCP tool registration detected. Tools are inferred from Flask HTTP routes, not registered via MCP protocol with schemas and descriptions. This violates the tool registration requirement and makes tool definitions invisible to MCP clients.
Parameter descriptions missing or trivial. 'context' parameter in update_context has only 'JSON object containing context data to be stored', no guidance on required fields, format, constraints, or examples. get_context has empty input schema with no documentation.
Tool descriptions are too short (<50 chars) and lack actionable context. 'Updates the global context data with JSON payload from POST request' does not explain WHEN to use this tool, what structure the context should have, or what happens on error. 'Retrieves the current global context data' is equally vague.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 23 | 2026-07-28+ | v2 |
No output schemas documented. LLMs cannot plan downstream calls or extract required fields. What does update_context return? What fields does get_context return? Undocumented outputs force LLMs to guess.
No error handling guidance. If update_context receives invalid JSON or get_context fails, what should the LLM do? No recovery hints, error classification, or actionable messages provided.