Loading ToolBench
A Tauri desktop application for visualizing and managing AI skills, plugins, MCPs, and workflows with a visual node-based editor
This server has critical definition gaps. Tool definitions are inferred from TypeScript/Rust code rather than explicit MCP registration. No input schemas are visible in the source, only parameter names and types are mentioned in comments. Descriptions exist but are minimal (10-50 chars), lacking LLM-optimized guidance on WHEN to use each tool or what happens on failure. No output schemas documented. No error handling guidance. The codebase is a Tauri desktop app (Next.js + Rust), not an MCP server, tool invocations use Tauri's `invoke()` API, not MCP protocol.
Adds a custom data source by mounting a directory path
Creates a new asset group
Deletes an asset group and moves its contents to trash
Deletes a skill and moves it to system trash, clearing telemetry records
Deletes a workflow file and moves it to trash
Gets the currently active/selected agent source
Retrieves all asset groups (skill groups, plugin groups, workflow groups)
No explicit MCP tool registration visible. Tool definitions are inferred from Tauri invoke() calls in TypeScript/Rust code, not declared via MCP protocol. This violates the tool registration pattern and makes the server non-compliant with MCP spec.
No input schemas visible in source code. Parameter types are mentioned in comments (e.g. 'sourceId: string') but no formal JSON Schema definitions are present.
Descriptions are too brief (10-50 chars). They state WHAT the tool does but lack WHEN to use it, prerequisites, or what happens on error. LLMs cannot distinguish between similar tools (e.g. get_sources vs get_active_source) with such minimal guidance.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 36 | 2026-07-28+ | v2 |
Retrieves all assets (skills, plugins, MCPs) from the active source
Gets the default directory for saving workflows
Retrieves all available agent sources (data sources for skills, plugins, workflows)
Initializes a demo pack of recommended skills for the active source
Loads a workflow from a file
Loads all saved workflows from the active source
Moves an asset to a different group
Reads the content of a file at the specified path
Removes a custom data source mount point
Saves a workflow with nodes and edges to a file
Switches to a different agent source by ID
No output schemas documented. Callers cannot know what fields to expect from responses (e.g. does get_sources return {id, name, path} or {sourceId, sourceName, sourcePath}?). This forces LLMs to guess field names for downstream tool calls.
No error handling guidance. Destructive tools (delete_skill, delete_group, delete_workflow) lack confirmation patterns or recovery instructions. No indication of which errors are retryable vs fatal.