Loading ToolBench
Keeps Roblox Studio's built-in MCP server connected, and tells you why it isn't when it's not.
Studio-mcp-doctor has 10 tools with inconsistent quality. Most tools have minimal descriptions (10-50 chars, well below the 194-char baseline). Input schemas are present but sparse: many tools have empty input objects (list_roblox_studios, get_studio_state, get_console_output, screen_capture, tools/list) with no parameters at all. Parameter descriptions exist for tools like search_game_tree and inspect_instance, but lack type constraints (no enums, ranges, or format specs). No output schemas are documented. Error handling is absent, no recovery guidance, no categorization of retryable vs fatal errors. The tool 'skill' is vague (what skill catalog? how does it relate to other tools?). Tool naming is mostly verb-noun compliant, but descriptions fail to explain WHEN to use each tool or what it returns.
Execute Luau code in Studio
Get console output from Studio
Get current Studio state
Fetch HTTP content with optional query and context
Inspect a specific instance by path
List connected Studio instances
Capture a screenshot of the Studio viewport
Six tools have empty input schemas (no parameters) with zero documentation of what they return. list_roblox_studios, get_studio_state, get_console_output, screen_capture, tools/list, and skill lack output schema documentation entirely.
Descriptions are uniformly too short (10-50 chars vs 194-char baseline). None explain WHEN to use the tool, what it returns, or how it chains with other tools. LLMs cannot determine tool selection intent from these descriptions.
Parameters lack type constraints. search_game_tree accepts max_depth and head_limit as integers with no min/max bounds. http_get accepts context_lines with no range. LLMs can pass absurd values (negative depths, 1M context lines) that break the tool.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 41 | 2026-07-28+ | v2 |
Search the game tree with optional path and depth filtering
Access skill catalog by name
List all available tools from Studio
Tool 'skill' is ambiguous. Name does not convey action (should be 'get_skill', 'list_skills', or 'access_skill_catalog'). Description 'Access skill catalog by name' does not explain what a skill is, when to call it, or what it returns.
No error handling guidance. Tools have no documented error responses, recovery paths, or categorization of retryable vs fatal errors. An LLM receiving a failure has no guidance on what to do next.