Loading ToolBench
Windows Forms chat interface for business users to interact with GitHub Copilot agents. Provides auto-update, chat history, loading indicators, error handling, and configurable settings for MCP server integration.
This PowerShell MCP server has severe definition quality gaps. Tool descriptions are present but generic (avg ~80 chars, below the 194-char baseline). Input schemas are visible but lack proper JSON Schema structure, parameters are documented as hashtables or objects without typed field definitions. No output schemas are documented. Critical issues: (1) tool names lack clear verb-noun patterns (e.g., 'Invoke-MCPQuery' is vague; 'Get-RealSalesReport' vs 'Get-MockSalesReport' creates confusion); (2) parameter descriptions are minimal and lack constraints (e.g., 'Period' accepts free-form strings with no enum); (3) no error handling guidance; (4) destructive tools (Cleanup-OldChatHistory, Update-AgentRepository) lack confirmation or dry-run patterns; (5) no output schema documentation for any tool. The server reads like a collection of utility functions rather than a production-grade agent toolkit.
Removes old chat history files exceeding the configured maximum history file limit
Returns default configuration object with MCP settings, UI configuration, update settings, agent settings, chat history settings, and logging configuration
Looks up mock customer information including tier, status, account history, lifetime value, and contact details
Retrieves mock inventory status for a product including stock levels, reserved units, available quantity, and reorder information
Generates mock sales report data with revenue, units, and margin information for a specified time period
Queries real MCP server for customer information lookup
Queries real MCP server for inventory status of a product
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields without knowing response structure.
Parameter schemas lack proper JSON Schema structure. 'Period' and 'ProductId' accept free-form strings with no enum constraints, inviting hallucinated values. 'Parameters' in Invoke-MCPQuery is a bare hashtable with no field definitions.
Destructive tools (Cleanup-OldChatHistory, Update-AgentRepository) lack confirmation, dry-run, or recovery guidance. No error handling tells LLM what to do on failure.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 35 | <=2025-11-25 | v2 |
Queries real MCP server for sales report data with revenue, units, and summary information for a specified period
Generic MCP server query function that sends REST requests to configured MCP server endpoints with optional Windows SSO authentication
Loads application configuration from config.json file with fallback to default configuration
Saves a chat message to the current session history file with timestamp and speaker information
Initializes a new chat session and creates a timestamped chat history file with session header
Performs git auto-update by fetching latest changes and pulling updates from configured remote and branch
Writes timestamped log messages to the application log file with configurable log level
Tool naming is inconsistent and ambiguous. 'Invoke-MCPQuery' is generic; 'Get-MockSalesReport' vs 'Get-RealSalesReport' creates confusion about which to use. 'Update-AgentRepository' does not clearly signal it performs git operations.
Parameter descriptions are minimal (avg ~30 chars, well below 72-char baseline). 'Period' lacks examples or valid values; 'StatusLabel' is a Windows Forms object with no guidance on how LLM should invoke it.