Loading ToolBench
A tray app for managing Claude Desktop profiles and other AI application profiles across macOS, Linux, and Windows
Agent Profiles has 7 tools with basic naming and descriptions, but significant gaps in schema completeness and parameter documentation. Tool names follow verb_noun convention (list_apps, add_profile, delete_profile), which is good. However, descriptions are generic and lack actionable context for LLM selection. Most critically: input schemas are present but parameter descriptions are minimal (1-2 sentences), and output schemas are entirely undocumented. No error handling guidance, no pagination support, and no indication of which operations are idempotent or safe to retry. The codebase is a Tauri desktop app (not an MCP server), so tool definitions appear inferred from TypeScript rather than explicitly registered via MCP protocol.
Creates a new profile for a specified application with a given label
Queries the operating system for the current autostart/login item state of the application
Deletes a profile and all its associated data, with safety checks to prevent deletion of running profiles
Lists all supported applications with their profiles and availability status
Calculates the total size in bytes of a profile directory, including all nested files and directories
Renames an existing profile with validation to prevent duplicates and blank labels
Enables or disables the application's login item in the operating system
No output schemas documented. LLMs cannot infer what list_apps, profile_size_bytes, or autostart_state return. This forces agents to guess field names and types, increasing errors.
Parameter descriptions are minimal (1-2 sentences) and lack actionable constraints. E.g., add_profile's 'label' param says 'max 80 characters, must be unique per app' but does not explain what happens if the limit is exceeded or uniqueness is violated.
delete_profile is destructive but has no confirmation step or dry-run mode. Agents can permanently delete profiles without a safety gate. No error guidance on what happens if a running profile is deleted.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 42 | 2026-07-28+ | v2 |
Tool definitions appear inferred from TypeScript source (src/main.ts) rather than explicitly registered via MCP protocol. No evidence of MCP server implementation, tool registration, or schema validation.
No error handling guidance. Tools do not document what errors are retryable, user-fixable, or fatal. E.g., add_profile fails if label is not unique, but the error response does not guide the LLM to retry with a different label.