Critical gaps across all dimensions. 52 tools lack visible input schemas in source code, cannot verify parameter types, descriptions, or constraints. Tool descriptions are present but minimal (10-50 chars typical), providing insufficient context for LLM selection. No evidence of output schema documentation. Error handling is absent, no recovery guidance, categorization, or actionable messages. Naming is inconsistent: verb_noun mostly present but some tools like 'existLink', 'massLink', 'pruneLongEdges' are unclear or non-standard. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite 15+ destructive operations. Security concerns: login/set-workspace/set-environment tools expose auth state management as tool parameters rather than server-side injection. Composition issues: 52 tools is excessive; many could be consolidated (e.g., createActor/createActors, getActor/searchActors, createAccount/updateAccount/deleteAccount). No pagination, no batch operation guidance, no chaining IDs in responses.

• Add explicit JSON Schema input definitions for all 52 tools. Include parameter types (string, number, boolean, array, object), required fields, enums for constrained values, and min/max for numeric ranges. Example: createActor should specify required fields (title, description), optional fields (status, domain), and return schema (actor_id, created_at).
• Expand tool descriptions to 100-200 chars following the pattern: WHAT (action), WHEN (use case), PREREQUISITES (required fields/auth), RETURNS (key fields). Example: 'Create a new actor (node) in a layer. Requires layer_id and actor title. Returns actor_id for use in createLink. Call after manageLayer to set up the layer.'
• Implement error handling with recovery guidance. For destructive tools (deleteActor, pruneLongEdges), return structured errors: {error_code, message, recovery_action}. Example: 'Actor not found (404). Did you mean: actor_123, actor_456? Call searchActors() to find the correct ID.'
• Move auth state management to server-side injection. Remove login, set-workspace, set-environment as tools. Instead, read SIMULATOR_AUTH_TOKEN, WORKSPACE_ID, SIMULATOR_PROFILE from environment or .env at startup. For stateless mode (per mcpserver.go Options.Stateless), accept auth via request context (apiclient.WithAuthorization).
• Consolidate overlapping tools: merge createActor + createActors into a single createActors(actors: array, mode: 'single'|'bulk'); merge getActor + searchActors into searchActors(query, limit, offset); merge createAccount/updateAccount/deleteAccount into manageAccount(action: 'create'|'update'|'delete', ...). Reduces tool count from 52 to ~35 and clarifies intent.