Loading ToolBench
Read-only local MCP server for explaining Microsoft Intune assignments.
Strong foundation with complete schemas, good descriptions, and proper error handling. All 6 tools have clear verb-noun naming (search_, get_, find_) and comprehensive input schemas with type definitions and parameter descriptions. Output schemas are well-documented with Zod validation. However, descriptions average ~100 chars (below the 194-char baseline for A+ tools) and lack explicit guidance on when to use each tool vs. alternatives. Parameter descriptions are present but terse. No critical gaps, but polish needed for LLM optimization.
Get the current Microsoft Graph connection status and authentication state.
Get all Intune policy assignments for a specific device, including configuration, compliance, and app policies.
Get all assignments for a specific Intune policy, including target groups, devices, and filter details.
Search for Intune device configuration and compliance policies by display name, category, or platform.
Find Intune policies that have no assignments, optionally filtered by category or platform.
Get all Intune policy assignments for a specific user, including configuration, compliance, and app policies.
Tool descriptions are terse (avg ~100 chars vs. 194-char baseline). Lack explicit guidance on when to use each tool vs. alternatives (e.g., when to call search_policies vs. unassigned_policies). LLMs need 50-200 char descriptions with WHEN/WHY context.
Parameter descriptions are minimal. E.g., 'nextCursor' appears in 5 tools but lacks explanation of pagination semantics or when to use it. 'platform' filter accepts 'windows', 'ios', 'android', 'macos' but these are not declared as enums, LLMs may hallucinate invalid values.
No explicit error recovery guidance in tool descriptions. E.g., if a policy_id is invalid, what should the LLM do? Retry? Call search_policies first? Error handling exists in code but is not surfaced in tool metadata.
Inferred effective spec: 2025-06-18+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 62 | 2025-06-18+ | v2 |
Tool composition: search_policies and unassigned_policies both return policy summaries but with different filtering logic. Descriptions do not clarify the distinction, LLMs may pick the wrong one. Add explicit 'Use this when...' guidance.
Output schemas are comprehensive but not documented in tool descriptions. LLMs cannot see the Zod schemas in code, they rely on descriptions to understand what fields to expect. Add brief output field summaries to descriptions.