Provider-agnostic Moodish concierge: Swiggy MCP-powered meal discovery and office lunch planning.
18 tools with basic schemas and descriptions, but significant gaps in parameter documentation and error handling. Most tools have descriptions (10-50 chars, below the 50-200 char baseline for LLM optimization). Parameter descriptions are minimal or missing context. No output schemas documented. No error recovery guidance. No tool annotations (readOnlyHint/destructiveHint). Naming is verb-forward but some tools combine multiple concerns (e.g., build_confirmed_cart does cart building + meal history recording). Security concern: aiApiKey and aiModel exposed as parameters in 3 tools instead of server-side injection.
Builds and confirms a shopping cart with selected items and add-ons
Cancels an active group meal session
Confirms and finalizes a group meal shopping cart
Creates a new group meal planning session
Retrieves the current state of a group meal session
Exports a user's taste memory and meal history
Interprets a meal-related message and returns structured meal preferences
Credentials exposed as tool parameters: aiApiKey and aiModel in interpret_meal_message, plan_personal_meal, plan_office_lunch. Should use server-side secret injection via environment variables.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields without knowing response structure.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 58 | 2026-07-28+ | v2 |
Plans an office lunch recommendation for a team based on team profile and preferences
Plans a personal meal recommendation based on user taste profile and recent meal history
Prepares a shopping cart for a meal recommendation
Prepares a shopping cart for a group meal session
Locks and ranks meal options for a group session
Records user feedback on a meal order
Selects a meal option for a group session
Submits meal preferences for a group meal session participant
Updates a user's taste profile with new preferences
Verifies access to a group meal session via invite passcode
Records a participant vote for a meal option in a group session
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Agents cannot distinguish safe reads from destructive writes without explicit hints.
Parameter descriptions lack actionable constraints. E.g., 'User identifier hash' does not explain format, length, or how to obtain it. Should include: 'SHA-256 hash of user email (64 hex chars)'.
No error recovery guidance. Tools do not return actionable error messages. E.g., 'Recommendation not found' should suggest 'Call plan_personal_meal() first to generate a recommendation.'