A desktop application built with Wails v3 that integrates MCP servers, skills, plugins, and agents for AI-powered workflows
Critical failures across all dimensions. Tool definitions are inferred from a React frontend component (ToolCard.tsx), not from actual MCP server code. No explicit tool registration visible in the Go backend. Input schemas are present but severely incomplete: most tools lack parameter type definitions (e.g., apply_patch, update_plan, plugin_list have empty input objects). Descriptions are minimal (10-40 chars), far below the 50-200 char baseline for LLM optimization. No output schemas documented. No error handling guidance. No parameter validation rules. The server appears to be a desktop UI wrapper around a Wails v3 application, not a proper MCP server with structured tool definitions.
Apply a patch to files
Ask the user a question
Create a new agent
Delegate a task to another agent
Check the status of a delegation
List available delegation targets
Execute a shell command
Execute commands in a persistent session
Tool definitions inferred from React component, not from actual MCP server code. No explicit tool registration visible in Go backend. Cannot verify actual implementation against schema.
7 tools have completely empty input schemas (apply_patch, update_plan, plugin_list, delegation_targets, and others).
All 31 tool descriptions are 10-40 characters, far below the 50-200 char baseline. Descriptions lack WHAT the tool does, WHEN to use it, and prerequisites. LLMs cannot determine selection criteria.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields. No pagination support visible for list tools.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 39 | 2026-07-28+ | v2 |
Generate an image from a prompt
Generate a video from a prompt
Find files matching a glob pattern
Search for patterns in files
List directory contents
Install a plugin
List installed plugins
Update a plugin
Read file contents
Request permissions from the user
Create a new skill
Install a skill
Modify an existing skill
Read skill information
Search for skills
Search for available tools
Unregister an agent
Update an existing agent
Update the execution plan
View an image file
Fetch content from a URL
Search the web
Write content to a file
Destructive tools (write_file, exec_command, exec_session, delete operations) lack error handling guidance, confirmation patterns, or dry-run support. No recovery paths documented.